Loading...
HomeMy WebLinkAbout03/17/2026 - Work Study - Packet171. Meeting Location: Contact us: Council Phone (360) portor hard 4407 216 Prospect Street cityhall@portorchardwa.gov Port Orchard, WA 98366 www.portorchardwa.gov City Council Work Study Session Tuesday, March 17, 2026 6:30 PM Pursuant to the Open Public Meetings Act, Chapter 42.30 RCW, the City Council is conducting its public meeting in a hybrid format with options for in -person attendance in the Council Chambers at City Hall or remote viewing and participation via Zoom (link below). The meeting is streamed live on the City's YouTube channel, click here. Remote Access Link: https://us02web.zoom.us/j/81994912407 Zoom Meeting ID: 819 9491 2407 Zoom Call -In: 1 253 215 8782 Guiding Principles Are we raising the bar in all of our actions? Are we honoring the past, but not living in the past? Are we building positive connections with our community and outside partners? Is the decision -making process building a diverse, equitable, and inclusive community? 1. Call to Order A. Pledge of Allegiance 2. Discussion Items A. Al Policy (Crocker/Dunham) Estimated Time: 15 Minutes B. Naming Rights and Philanthropic Giving (Mayor) Estimated Time: 15 Minutes C. Town Hall Meeting Format Estimated Time: 30 Minutes 3. Good of the Order 4. Adjournment ADA Requirements March 17, 2026 Meeting Agenda In compliance with the American with Disabilities Act, if you need accommodations to participate in this meeting, please contact the City Clerk's office at (360) 876-4407. Notification at least 48 hours in advance of meeting will enable the City to make arrangements to assure accessibility to this meeting. REMINDER: Please silence all electronic devices while City Council is in session. To subscribe to our general news & public notices click the link: http://portorchardwa.gov/subscribe For current City Council member and contact information, please visit https://portorchardwa.gov/departments/city- council/. For Committee Membership please visit https://portorchardwa.gov/city-council-advisory-committees/. March 17, 2026 Meeting Agenda 2 Ong City of Port Orchard ORCHARD 216 Prospect Street, Port Orchard, WA 98366 (360) 876-4407 • FAX (360) 895-9029 Agenda Staff Report Discussion Items: Al Policy (Crocker/Dunham) Estimated Time: 15 Minutes Meeting Date: March 17, 2026 Prepared By: Sean Dunham, IT Manager Summary and Background: As Artificial Intelligence tools become integrated into standard municipal software (e.g., Microsoft 365 Copilot, cybersecurity monitoring, and automated utility analytics), it is necessary for the City of Port Orchard to establish a formal policy. This policy ensures that City employees use Al in a manner that is transparent, ethical, and secure. The framework aligns with the Washington State Al Task Force 2026 Recommendations and the NIST Al Risk Management Framework, focusing on protecting sensitive data while leveraging technology to improve operational efficiency. In early 2025, the City established an internal Al Committee to evaluate the impact of generative Al and automated systems on City operations. Since then, several factors have necessitated a formal policy: State Compliance: The Washington State Legislature is currently reviewing SB 5984 (2026), which requires specific disclosures when interacting with Al systems. Operational Integration: Productivity tools used by City staff now include embedded Al features that require "opt -in" governance to prevent the accidental upload of protected or sensitive municipal data to public models. Risk Management: Without a policy, the City faces risks related to "model drift" (inaccurate data output), algorithmic bias, and potential cybersecurity vulnerabilities. The draft policy is built upon four primary pillars: Data Privacy: Strict prohibition on inputting PII (Personally Identifiable Information) or sensitive City records into public Al models. Human Oversight: "Human -in -the -loop" requirement: all Al generated content (reports, code, or 3 public notices) must be reviewed and verified by a staff member. Transparency: Public facing Al tools (e.g., chatbots) must be clearly labeled as such to comply with state transparency standards. Risk Classification: Systems are categorized as Low Risk (internal drafting) or High Risk (systems affecting public health, safety, or legal rights), with the latter requiring Council approval. Relationship to Comprenhensive Plan: N/A Recommendation: Staff recommends that the City Council review the draft Al Usage Policy and provide direction to the Al Committee for finalization prior to formal adoption at a regular Council meeting. Motion for Consideration: Staff recommends that the City Council discuss the balance between innovation and risk mitigation, specifically regarding the use of Al in public facing communications. Has item been presented to Committee/Work Study? If so, which one: Finance Committee Fiscal Impact: There is no immediate fiscal impact associated with the adoption of this policy. However, future procurement of Al -specific tools or enhanced cybersecurity monitoring may require budgetary considerations in the 2027-2028 Biennial Budget. Alternatives: No Discussion Attachments: 910 Port Orchard Al Policy-FINALv2.pdf NIST.Al.100-1.pdf Port Orchard Acquisition of Al Review Guidelines.pdf Port Orchard Al Vendor Factsheet.pdf Port Orchard Algorithmic Impact Assessment (AIA) Form.pdf Port Orchard Data Classification Guidelines.pdf Port_Orchard_AI_Committee_Pol icy. pptx 4 CITY OF PORT ORCHARD 910 - Artificial Intelligence Policy Effective Date: ASAP PURPOSE The purpose of this policy is to set forth the requirements to observe when acquiring and using software that meets the definition of "artificial intelligence." SCOPE All individuals and entities (herein defined as "Users"), including City departments, employees, elected officials, vendors, contractors, and volunteers, who operate under the authority of the City of Port Orchard and engage with City data are bound by this policy. DEFINITIONS Terms used in the current Artificial Intelligence space are fluid and dynamic. Attached in Appendix A is a list of current definitions that have been approved by the Al committee to ensure a shared understanding of its scope and application. ARTIFICIAL INTELLIGENCE (Al) PRINCIPLES These Principles describe general codes of conduct that represent the values and responsibilities of the City to its residents. This policy serves to inform Users in their use of Al technology. Users shall adhere to the principles and requirements outlined in this policy. • Innovation: The City values public service innovation to meet our residents' needs. We commit to responsibly explore and continuously evaluate Al technologies, which will improve our services and advance beneficial outcomes for our community. • Transparency and Accountability: The City values transparency and accountability and understands the importance of these values in our use of Al systems. The City will ensure that the development, use, and deployment of Al systems are evaluated for and compliant with all laws and regulations applicable to the City prior to use and will make documentation related to the use of Al systems available publicly. • Validity and Reliability: The City will work to ensure that Al systems perform reliably and consistently under the conditions of expected use, and that ongoing evaluation of system accuracy throughout the development and/or deployment lifecycle is managed, governed, and auditable. • Bias, Harm Reduction, and Equity: The City acknowledges that Al systems have the potential to perpetuate inequity and bias resulting in unintended and potentially harmful outcomes. The City will evaluate Al systems with a strong focus on equity, addressing potential impacts arising from data, human, or algorithmic bias. • Data Privacy: The City values data privacy and understands the importance of protecting personal data. The City strives to ensure that policies and standard operating procedures reduce privacy risks, and are applied to Al systems throughout development, testing, deployment, and use. • Explainability and Interpretability: The City understands the importance of leveraging Al systems, models, and outputs that are easily interpreted and explained. The City will attempt to ensure all Al systems utilized, and their outputs, are communicated in clear language, representative of the context in which they are deployed. • Security and Resiliency: Securing our data, systems, and infrastructure is important to the City. The City will ensure Al systems are evaluated for resilience and can maintain confidentiality, integrity, and availability of data for critical City systems. The City will actively work to minimize security risks in alignment with governing policy and identified best practices. POLICY 1. Acquisition and Usage of Al Technology 1.1. The City has a list of approved Al tools that can be used in accordance with individual department policy (see Appendix B). This list applies to all employees of the City and elected officials. 1.2. Consistent with the City's standards for Acquisition of Technology Resources, Department directors may request acquisition of Al tools (not listed in Appendix B) through the City's current IT request process. Departments may not acquire or use Al systems without following the approval process identified in this policy. 1.3. The IT Department shall review requests according to its current risk and impact methodology, which shall include specific review criteria for Al technology (see Port Orchard Acquisition of Al Review Guidelines). The IT Department will then bring the request to the Al Committee for approval or denial. 1.4. The City's standard for technology acquisition applies to all technology, including Open Source, Free to Use software, or SaaS (Software as a Service) tools. 1.5. If a technology that has already been approved for use in the City adds or incorporates Al capabilities, the IT Department shall immediately issue a technical restriction to suspend the use of those new Al capabilities pending formal review. The Al Committee will immediately be notified and will then evaluate the change to ensure it continues to follow this policy, according to its current risk and impact methodology (see Section 1.3). If the Al Committee approves the new Al capabilities, the IT Department may lift the technical restriction. 1.6. The City's IT Department shall revoke authorization for a technology that adds Al capabilities and/or restrict the use of those Al capabilities; if those Al capabilities present risks that cannot be effectively mitigated to comply with this policy or other City policies until the risks can be addressed. • The IT Department's responsibilities include the technical authority to implement immediate and mandatory suspension or restriction of new Al capabilities added to existing software, as required by Section 1.5. The IT Department is responsible for notifying the impacted department(s) of the suspension and/or restrictions. 2. Use of Al Outputs 6 2.1. Outputs of Al systems shall be reviewed by human(s) prior to each use in an official City capacity ("Human in the Loop" or HITL). 3. Attribution, Accountability, and Transparency of Authorship 3.1. Any outputs generated by an Al system, and used substantively in a final product, requires attribution to the relevant Al system used. • Departments shall interpret substantive use thresholds to be consistent with the principles outlined in this document, as well as relevant intellectual property laws. • All attributions should include the name of the Al system used plus an HITL assertion (which should include the department or group that reviewed/edited the content). **Example** "Some material in this brochure was generated using ChatGPT 4.0 and was reviewed for accuracy by a member of the Department of Human Resources before publication." 4. Reducing Bias and Harm 4.1. Al systems may produce outputs based on stereotypes or use data that is biased. Users will review and evaluate Al generated content and the designated HITL reviewer shall ensure that the output is accurate and free of discrimination and bias (HITL). 5. Data Privacy 5.1. Use of Al tools shall be consistent with the principles and standards described in the City's Data Privacy Policy, and Information Security Policy. 5.2. Unless suitable enterprise controls and data protection mitigations are in place, users shall not submit any of the following to Al systems outside of the City's control: • Data classified by the City's data classification guidelines as "Confidential" or "Highly Confidential" as listed in the Port Orchard Data Classification Guidelines. • Data that constitutes a Protected Data or Public Record that is exempt from disclosure under the Revised Code of Washington (RCW) (RCW 42.56, et al.) or other applicable laws. • Any other non public information such as: preliminary drafts, intra-agency memorandums, or sensitive operational data that is not officially approved for public disclosure. • Data that is not considered to be acceptable to disclose to the public 5.3. No City data or records, including inputs or prompts, are to be used for training or parameter tuning for Al models outside the City's control. Al technologies that cannot prevent City data or records from contributing to their language models may not be used. 6. Public Records & City Records Management 6.1. All records generated, used, or stored by Al vendors or solutions may be considered public records and must be disclosed upon request. 7 6.2. All Al solutions and/or vendors approved for City use shall be required to support retrieval and export of all prompts and outputs (either via exposed functionality or through vendor contract assurances). 6.3. Users who use Al tools are required to maintain, or be able to retrieve upon request, records of inputs, prompts, and outputs in a manner consistent with the City's records management and public disclosure policies and practices. EXCEPTIONS Any exceptions to this policy must be approved in advance through submission to the City Al Committee. POLICY COMPLIANCE Noncompliance may result in the Mayor, department directors, or their designees, imposing disciplinary action, up to and including termination of employment or vendor contract. RELATED STANDARDS AND POLICIES • 901 - City of Port Orchard IT Data Security Policy • 902 - City of Port Orchard IT Data Privacy Policy • City of Port Orchard Acquisition of Al Review Guidelines • City of Port Orchard Data Classification Guidelines • National Institute of Standards and Technology Artificial Intelligence Risk Management Framework (NIST Al 100-1) (In Draft Folder) • Revised Code of Washington (RCW): o Ethics In Public Service (RCW 42.52) o Public Records Act (RCW 42.56) o Preservation and Destruction of Public Records (RCW 40.14) 8 RESPONSIBILITIES This policy will be maintained through the City's Al Committee. Their responsibilities include creating and maintaining the Al risk and impact criteria, developing mandatory standards for the uniform interpretation and measurement of "Substantive Use" (Section 3.1), criteria by which to evaluate suggestions for amendments to Appendix B, (see City of Port Orchard Al Review and Guidelines Policy), including the creation of mandatory departmental standards for designating and training HITL reviewers. DOCUMENT CONTROL This policy shall be effective on TBD and shall be reviewed regularly, or as appropriate. POLICY REVIEW AND APPROVAL HISTORY Version Content v 0.1 Initial Draft v 0.2 2nd Draft v 0.5 5th Draft v 0.6 6th Draft Contributors Reviewer(s): Noah Crocker -Finance Director, Rebecca Zick-Deputy Finance Director, Kori Pearson- Acct Asst III/IT Specialist, Alan Iwashita-POPD Deputy Chief, Sean Dunham -IT Manager, Jake Langston -IT Specialist, Caden Cucciardi- Asset Management Tech Reviewer(s): Noah Crocker -Finance Director, Rebecca Zick-Deputy Finance Director, Kori Pearson- Acct Asst III/IT Specialist, Alan Iwashita-POPD Deputy Chief, Sean Dunham -IT Manager, Jake Langston -IT Specialist, Caden Cucciardi- Asset Management Tech Reviewer(s): Noah Crocker -Finance Director, Rebecca Zick-Deputy Finance Director, Kori Pearson- Acct Asst III/IT Specialist, Debbie Lund — HR Director, Sean Dunham -IT Manager, Jake Langston -IT Specialist, Caden Cucciardi- Asset Management Tech, Jenine Floyd — Deputy City Clerk, Nick Bond — DCD Director Reviewer(s): Noah Crocker -Finance Director, Rebecca Zick-Deputy Finance Director, Kori Pearson -Acct Asst III/IT Specialist, Sean Dunham -IT Manager, Jake Langston -IT Specialist, Caden Cucciardi- Asset Management Tech Reviewer(s): Noah Crocker -Finance Director, Rebecca Zick-Deputy Finance Director, Kori Pearson- Acct Asst III/IT Specialist, Debbie Lund -HR Director, Sean Dunham -IT Manager, Jake Langston -IT Specialist, Caden Cucciardi- Asset Management Tech, Jenine Floyd —Deputy City Clerk, Nick Bond—DCD Director, Jim Fisk —Senior Planner, Alan Iwashita-Deputy Police Chief Reviewer(s): Noah Crocker -Finance Director, Kori Pearson- Acct Asst III/IT Specialist, Debbie Lund —HR Director, Sean Dunham -IT Manager, Jake Langston -IT Specialist, Caden Cucciardi- Asset Management Tech, Jenine Floyd —Deputy City Clerk, Nick Bond—DCD Director, Jim Fisk —Senior Planner, Alan Iwashita-Deputy Police Chief Approval Date July 17, 2025 August 21,2025 November 20, 2025 December 9, 2025 9 v 1.0 Final Version Reviewer(s): Noah Crocker -Finance Director, Rebecca Zick-Deputy Finance Director, Kori Pearson- Acct Asst III/IT Specialist, Debbie Lund -HR Director, Sean Dunham -IT Manager, Jake Langston -IT Specialist, Caden Cucciardi- Asset Management Tech, Brandy Wallace -City Clerk Decemberl 6, 2025 10 APPENDIX A: GENERAL Al DEFINITIONS Al (Artificial Intelligence): The capability of a machine or system to perform tasks that typically require human intelligence, such as generating text or audiovisual content, making or recommending decisions, analyzing data, or automating processes. Al Algorithm / Al Model: A set of programmed instructions that processes data to perform tasks, make decisions, or solve problems within an Al system. Al Assistant: An Al tool which is intended to aid a user in their day-to-day work by suggesting content, retrieving information, automating processes, and performing other similar tasks. Al Policy: A document that provides a framework for the effective and responsible use of Al systems within an organization. Al System: Any tool, software, process, and workflow, or other system which is based on Al technology, or which uses Al technology as a key component of that system. Al Tool: A piece of software which provides Al functionality and can be applied to a specific use case. Al Use Case: A specific task or purpose for which an Al tool is used or under consideration. Al User: An individual who is responsible for using, developing, purchasing, configuring, or maintaining Al systems. Al User Guide: A document which supplements an Al policy with more detailed guidance on how to implement the policy. Agentic Al: Autonomous Al systems that can act independently to achieve predefined goals, making decisions and taking actions without constant human oversight. Anonymization: A process by which data is altered so that it cannot be connected to specific individuals or organizations. Bias: Systematic tendencies that can exist within Al systems, often stemming from flawed data, algorithms, or design processes, that may lead to discriminatory or inaccurate outcomes affecting certain groups or individuals. Black Box Algorithm: An Al algorithm which produces decisions or other outputs with little or no mechanism for the user to analyze the logic which led to that result. Data Privacy: The protection of non-public information about a person or organization from disclosure without their consent. Data Output: Refers to the information produced by a computer or device as a result of processing input data. It can take various forms, including; Text, Images, Audio, Video, and more. Generative Al: Al systems which use algorithms to create text, audio, image, or video content based on some combination of user prompts and stored data and instructions. Human -in -the -Loop (HITL): A system design philosophy that requires human oversight and intervention in Al -driven processes. This ensures that final, critical decisions are made by a person, not the algorithm. Large Language Model (LLM): A type of generative Al that is trained on a massive amount of text data to understand, summarize, generate, and predict new human -like language. Machine Learning: A type of Al which uses algorithms to extract information from and recognize patterns in data, often used for forecasting, prediction, classification, or analysis. Open Data: Information derived from an organization's operations which is shared publicly to promote transparency or for use by external parties. Protected Data: Information generated or acquired in the course of an organization's operations which is not intended or approved for public disclosure. 11 Public Record: Records stemming from an organization's operations which must by law be shared upon request, whether or not it is actively published as open data. 12 APPENDIX B: Al TECHNOLOGY REGISTRY (Al Tools Currently Approved By Al Committee) Microsoft Copilot 13 NIST AI 100-1 -t Artificial Intelligence Risk Management Framework (AI RMF 1.0) N NATIONAL INSTITUTE OF LST" I STANDARDS AND TECHNOLOGY U.S. DEPARTMENT OF COMMERCE NIST AI 100-1 Artificial Intelligence Risk Management Framework (AI RMF 1.0) This publication is available free of charge from: https://doi.org/10.6028/NIST.AI. 100-1 January 2023 I 443 U.S. Department of Commerce Gina M. Raimondo, Secretary National Institute of Standards and Technology Laurie E. Locascio, NIST Director and Under Secretary of Commerce for Standards and Technology 15 Certain commercial entities, equipment, or materials may be identified in this document in order to describe an experimental procedure or concept adequately. Such identification is not intended to imply recommenda- tion or endorsement by the National Institute of Standards and Technology, nor is it intended to imply that the entities, materials, or equipment are necessarily the best available for the purpose. This publication is available free of charge from: https://doi.org/10.6028/NIST.AI.100-1 Update Schedule and Versions The Artificial Intelligence Risk Management Framework (AI RMF) is intended to be a living document. NIST will review the content and usefulness of the Framework regularly to determine if an update is appro- priate; a review with formal input from the AI community is expected to take place no later than 2028. The Framework will employ a two -number versioning system to track and identify major and minor changes. The first number will represent the generation of the AI RMF and its companion documents (e.g., 1.0) and will change only with major revisions. Minor revisions will be tracked using ".n" after the generation number (e.g., 1.1). All changes will be tracked using a Version Control Table which identifies the history, including version number, date of change, and description of change. NIST plans to update the Al RMF Playbook frequently. Comments on the AI RMF Playbook may be sent via email to AIframework@nist.gov at any time and will be reviewed and integrated on a semi-annual basis. 16 Table of Contents Executive Summary 1 Part 1: Foundational Information 4 1 Framing Risk 4 1.1 Understanding and Addressing Risks, Impacts, and Harms 4 1.2 Challenges for AI Risk Management 5 1.2.1 Risk Measurement 5 1.2.2 Risk Tolerance 7 1.2.3 Risk Prioritization 7 1.2.4 Organizational Integration and Management of Risk 8 2 Audience 9 3 AI Risks and Trustworthiness 12 3.1 Valid and Reliable 13 3.2 Safe 14 3.3 Secure and Resilient 15 3.4 Accountable and Transparent 15 3.5 Explainable and Interpretable 16 3.6 Privacy -Enhanced 17 3.7 Fair — with Harmful Bias Managed 17 4 Effectiveness of the AI RMF 19 Part 2: Core and Profiles 20 5 AI RMF Core 20 5.1 Govern 21 5.2 Map 24 5.3 Measure 28 5.4 Manage 31 6 AI RMF Profiles 33 Appendix A: Descriptions of AI Actor Tasks from Figures 2 and 3 35 Appendix B: How AI Risks Differ from Traditional Software Risks 38 Appendix C: AI Risk Management and Human -Al Interaction 40 Appendix D: Attributes of the AI RMF 42 List of Tables Table 1 Categories and subcategories for the GOVERN function. 22 Table 2 Categories and subcategories for the MAP function. 26 Table 3 Categories and subcategories for the MEASURE function. 29 Table 4 Categories and subcategories for the MANAGE function. 32 i 17 NIST AI 100-1 AI RMF 1.0 List of Figures Fig. 1 Examples of potential harms related to AI systems. Trustworthy AI systems and their responsible use can mitigate negative risks and contribute to bene- fits for people, organizations, and ecosystems. 5 Fig. 2 Lifecycle and Key Dimensions of an AI System. Modified from OECD (2022) OECD Framework for the Classification of AI systems — OECD Digital Economy Papers. The two inner circles show AI systems' key di- mensions and the outer circle shows AI lifecycle stages. Ideally, risk man- agement efforts start with the Plan and Design function in the application context and are performed throughout the AI system lifecycle. See Figure 3 for representative AI actors. 10 Fig. 3 AI actors across AI lifecycle stages. See Appendix A for detailed descrip- tions of AI actor tasks, including details about testing, evaluation, verifica- tion, and validation tasks. Note that AI actors in the AI Model dimension (Figure 2) are separated as a best practice, with those building and using the models separated from those verifying and validating the models. 11 Fig. 4 Characteristics of trustworthy AI systems. Valid & Reliable is a necessary condition of trustworthiness and is shown as the base for other trustworthi- ness characteristics. Accountable & Transparent is shown as a vertical box because it relates to all other characteristics. 12 Fig. 5 Functions organize AI risk management activities at their highest level to govern, map, measure, and manage AI risks. Governance is designed to be a cross -cutting function to inform and be infused throughout the other three functions. 20 Page ii 18 NIST Al 100-1 AI RMF 1.0 Executive Summary Artificial intelligence (AI) technologies have significant potential to transform society and people's lives — from commerce and health to transportation and cybersecurity to the envi- ronment and our planet. AI technologies can drive inclusive economic growth and support scientific advancements that improve the conditions of our world. AI technologies, how- ever, also pose risks that can negatively impact individuals, groups, organizations, commu- nities, society, the environment, and the planet. Like risks for other types of technology, AI risks can emerge in a variety of ways and can be characterized as long- or short-term, high - or low -probability, systemic or localized, and high- or low -impact. The AI RMF refers to an Al system as an engineered or machine -based system that can, for a given set of objectives, generate outputs such as predictions, recommenda- tions, or decisions influencing real or virtual environments. AI systems are designed to operate with varying levels of autonomy (Adapted from: OECD Recommendation on AI:2019; ISO/IEc 22989:2022). While there are myriad standards and best practices to help organizations mitigate the risks of traditional software or information -based systems, the risks posed by AI systems are in many ways unique (See Appendix B). AI systems, for example, may be trained on data that can change over time, sometimes significantly and unexpectedly, affecting system function- ality and trustworthiness in ways that are hard to understand. AI systems and the contexts in which they are deployed are frequently complex, making it difficult to detect and respond to failures when they occur. AI systems are inherently socio-technical in nature, meaning they are influenced by societal dynamics and human behavior. AI risks — and benefits — can emerge from the interplay of technical aspects combined with societal factors related to how a system is used, its interactions with other AI systems, who operates it, and the social context in which it is deployed. These risks make AI a uniquely challenging technology to deploy and utilize both for orga- nizations and within society. Without proper controls, AI systems can amplify, perpetuate, or exacerbate inequitable or undesirable outcomes for individuals and communities. With proper controls, AI systems can mitigate and manage inequitable outcomes. AI risk management is a key component of responsible development and use of AI sys- tems. Responsible AI practices can help align the decisions about AI system design, de- velopment, and uses with intended aim and values. Core concepts in responsible AI em- phasize human centricity, social responsibility, and sustainability. AI risk management can drive responsible uses and practices by prompting organizations and their internal teams who design, develop, and deploy Al to think more critically about context and potential or unexpected negative and positive impacts. Understanding and managing the risks of AI systems will help to enhance trustworthiness, and in turn, cultivate public trust. Page 1 19 NIST Al 100-1 AI RMF 1.0 Social responsibility can refer to the organization's responsibility "for the impacts of its decisions and activities on society and the environment through transparent and ethical behavior" (ISO 26000:2010). Sustainability refers to the "state of the global system, including environmental, social, and economic aspects, in which the needs of the present are met without compromising the ability of future generations to meet their own needs" (Iso/IEc TR 24368:2022). Responsible AI is meant to result in technology that is also equitable and accountable. The expectation is that organizational practices are carried out in accord with "professional responsibility," defined by ISO as an approach that "aims to ensure that professionals who design, develop, or deploy AI systems and applications or AI -based products or systems, recognize their unique position to exert influence on people, society, and the future of AI" (Iso/IEc TR 24368:2022). As directed by the National Artificial Intelligence Initiative Act of 2020 (P.L. 116-283), the goal of the AI RMF is to offer a resource to the organizations designing, developing, deploying, or using AI systems to help manage the many risks of AI and promote trustwor- thy and responsible development and use of AI systems. The Framework is intended to be voluntary, rights -preserving, non -sector -specific, and use -case agnostic, providing flexibil- ity to organizations of all sizes and in all sectors and throughout society to implement the approaches in the Framework. The Framework is designed to equip organizations and individuals — referred to here as Al actors — with approaches that increase the trustworthiness of AI systems, and to help foster the responsible design, development, deployment, and use of AI systems over time. AI actors are defined by the Organisation for Economic Co-operation and Development (OECD) as "those who play an active role in the AI system lifecycle, including organiza- tions and individuals that deploy or operate AI" [OECD (2019) Artificial Intelligence in Society —OECD iLibrary] (See Appendix A). The AI RMF is intended to be practical, to adapt to the AI landscape as AI technologies continue to develop, and to be operationalized by organizations in varying degrees and capacities so society can benefit from AI while also being protected from its potential harms. The Framework and supporting resources will be updated, expanded, and improved based on evolving technology, the standards landscape around the world, and AI community ex- perience and feedback. NIST will continue to align the AI RMF and related guidance with applicable international standards, guidelines, and practices. As the AI RMF is put into use, additional lessons will be learned to inform future updates and additional resources. The Framework is divided into two parts. Part 1 discusses how organizations can frame the risks related to AI and describes the intended audience. Next, AI risks and trustworthi- ness are analyzed, outlining the characteristics of trustworthy AI systems, which include Page 2 20 NIST Al 100-1 AI RMF 1.0 valid and reliable, safe, secure and resilient, accountable and transparent, explainable and interpretable, privacy enhanced, and fair with their harmful biases managed. Part 2 comprises the "Core" of the Framework. It describes four specific functions to help organizations address the risks of AI systems in practice. These functions - GOVERN, MAP, MEASURE, and MANAGE — are broken down further into categories and subcate- gories. While GOVERN applies to all stages of organizations' AI risk management pro- cesses and procedures, the MAP, MEASURE, and MANAGE functions can be applied in AI system -specific contexts and at specific stages of the AI lifecycle. Additional resources related to the Framework are included in the AI RMF Playbook, which is available via the NIST AI RMF website: https://www.nist.gov/itl/ai-risk-management-framework. Development of the AI RMF by NIST in collaboration with the private and public sec- tors is directed and consistent with its broader AI efforts called for by the National AI Initiative Act of 2020, the National Security Commission on Artificial Intelligence recom- mendations, and the Plan for Federal Engagement in Developing Technical Standards and Related Tools. Engagement with the AI community during this Framework's development — via responses to a formal Request for Information, three widely attended workshops, public comments on a concept paper and two drafts of the Framework, discussions at mul- tiple public forums, and many small group meetings — has informed development of the AI RMF 1.0 as well as AI research and development and evaluation conducted by NIST and others. Priority research and additional guidance that will enhance this Framework will be captured in an associated AI Risk Management Framework Roadmap to which NIST and the broader community can contribute. Page 3 21 NIST AI 100-1 AI RMF 1.0 Part 1: Foundational Information 1. Framing Risk AI risk management offers a path to minimize potential negative impacts of AI systems, such as threats to civil liberties and rights, while also providing opportunities to maximize positive impacts. Addressing, documenting, and managing AI risks and potential negative impacts effectively can lead to more trustworthy AI systems. 1.1 Understanding and Addressing Risks, Impacts, and Harms In the context of the AI RMF, risk refers to the composite measure of an event's probability of occurring and the magnitude or degree of the consequences of the corresponding event. The impacts, or consequences, of AI systems can be positive, negative, or both and can result in opportunities or threats (Adapted from: Iso 31000:2018). When considering the negative impact of a potential event, risk is a function of 1) the negative impact, or magni- tude of harm, that would arise if the circumstance or event occurs and 2) the likelihood of occurrence (Adapted from: OMB Circular A-130:2016). Negative impact or harm can be experienced by individuals, groups, communities, organizations, society, the environment, and the planet. "Risk management refers to coordinated activities to direct and control an organiza- tion with regard to risk" (Source: Iso 31000:2018). While risk management processes generally address negative impacts, this Framework of- fers approaches to minimize anticipated negative impacts of AI systems and identify op- portunities to maximize positive impacts. Effectively managing the risk of potential harms could lead to more trustworthy AI systems and unleash potential benefits to people (individ- uals, communities, and society), organizations, and systems/ecosystems. Risk management can enable AI developers and users to understand impacts and account for the inherent lim- itations and uncertainties in their models and systems, which in turn can improve overall system performance and trustworthiness and the likelihood that AI technologies will be used in ways that are beneficial. The AI RMF is designed to address new risks as they emerge. This flexibility is particularly important where impacts are not easily foreseeable and applications are evolving. While some AI risks and benefits are well-known, it can be challenging to assess negative impacts and the degree of harms. Figure 1 provides examples of potential harms that can be related to AI systems. AI risk management efforts should consider that humans may assume that Al systems work — and work well — in all settings. For example, whether correct or not, AI systems are often perceived as being more objective than humans or as offering greater capabilities than general software. Page 4 22 MST Al 100-1 AI RMF 1.0 .:niizir- • Individual: Harm to a person's • Harm to an organization's • Harm to interconnected and civil liberties, rights, physical or business operations. interdependent elements and psychological safety, or economic resources. opportunity. • Group/Community: Harm to a • Harm to an organization from • Harm to the global financial group such as discrimination security breaches or monetary system, supply chain, or against a population sub -group. loss, interrelated systems. • Societal: Harm to democratic • Harm to an organization's • Harm to natural resources, the participation or educational reputation. environment, and planet. access. Fig. 1. Examples of potential harms related to AI systems. Trustworthy AI systems and their responsible use can mitigate negative risks and contribute to benefits for people, organizations, and ecosystems. 1.2 Challenges for AI Risk Management Several challenges are described below. They should be taken into account when managing risks in pursuit of AI trustworthiness. 1.2.1 Risk Measurement AI risks or failures that are not well-defined or adequately understood are difficult to mea- sure quantitatively or qualitatively. The inability to appropriately measure AI risks does not imply that an AI system necessarily poses either a high or low risk. Some risk measurement challenges include: Risks related to third -party software, hardware, and data: Third -party data or systems can accelerate research and development and facilitate technology transition. They also may complicate risk measurement. Risk can emerge both from third -party data, software or hardware itself and how it is used. Risk metrics or methodologies used by the organization developing the AI system may not align with the risk metrics or methodologies uses by the organization deploying or operating the system. Also, the organization developing the AI system may not be transparent about the risk metrics or methodologies it used. Risk measurement and management can be complicated by how customers use or integrate third - party data or systems into AI products or services, particularly without sufficient internal governance structures and technical safeguards. Regardless, all parties and AI actors should manage risk in the AI systems they develop, deploy, or use as standalone or integrated components. Tracking emergent risks: Organizations' risk management efforts will be enhanced by identifying and tracking emergent risks and considering techniques for measuring them. Page 5 23 NIST Al 100-1 AI RMF 1.0 AI system impact assessment approaches can help AI actors understand potential impacts or harms within specific contexts. Availability of reliable metrics: The current lack of consensus on robust and verifiable measurement methods for risk and trustworthiness, and applicability to different AI use cases, is an AI risk measurement challenge. Potential pitfalls when seeking to measure negative risk or harms include the reality that development of metrics is often an institu- tional endeavor and may inadvertently reflect factors unrelated to the underlying impact. In addition, measurement approaches can be oversimplified, gamed, lack critical nuance, be- come relied upon in unexpected ways, or fail to account for differences in affected groups and contexts. Approaches for measuring impacts on a population work best if they recognize that contexts matter, that harms may affect varied groups or sub -groups differently, and that communities or other sub -groups who may be harmed are not always direct users of a system. Risk at different stages of the AI lifecycle: Measuring risk at an earlier stage in the AI lifecycle may yield different results than measuring risk at a later stage; some risks may be latent at a given point in time and may increase as AI systems adapt and evolve. Fur- thermore, different AI actors across the AI lifecycle can have different risk perspectives. For example, an AI developer who makes AI software available, such as pre -trained mod- els, can have a different risk perspective than an AI actor who is responsible for deploying that pre -trained model in a specific use case. Such deployers may not recognize that their particular uses could entail risks which differ from those perceived by the initial developer. All involved AI actors share responsibilities for designing, developing, and deploying a trustworthy AI system that is fit for purpose. Risk in real -world settings: While measuring AI risks in a laboratory or a controlled environment may yield important insights pre -deployment, these measurements may differ from risks that emerge in operational, real -world settings. Inscrutability: Inscrutable AI systems can complicate risk measurement. Inscrutability can be a result of the opaque nature of AI systems (limited explainability or interpretabil- ity), lack of transparency or documentation in AI system development or deployment, or inherent uncertainties in AI systems. Human baseline: Risk management of AI systems that are intended to augment or replace human activity, for example decision making, requires some form of baseline metrics for comparison. This is difficult to systematize since AI systems carry out different tasks — and perform tasks differently — than humans. Page 6 24 MST Al 100-1 AI RMF 1.0 1.2.2 Risk Tolerance While the AI RMF can be used to prioritize risk, it does not prescribe risk tolerance. Risk tolerance refers to the organization's or AI actor's (see Appendix A) readiness to bear the risk in order to achieve its objectives. Risk tolerance can be influenced by legal or regula- tory requirements (Adapted from: ISO GUIDE 73). Risk tolerance and the level of risk that is acceptable to organizations or society are highly contextual and application and use -case specific. Risk tolerances can be influenced by policies and norms established by AI sys- tem owners, organizations, industries, communities, or policy makers. Risk tolerances are likely to change over time as AI systems, policies, and norms evolve. Different organiza- tions may have varied risk tolerances due to their particular organizational priorities and resource considerations. Emerging knowledge and methods to better inform harm/cost-benefit tradeoffs will con- tinue to be developed and debated by businesses, governments, academia, and civil society. To the extent that challenges for specifying AI risk tolerances remain unresolved, there may be contexts where a risk management framework is not yet readily applicable for mitigating negative AI risks. The Framework is intended to be flexible and to augment existing risk practices which should align with applicable laws, regulations, and norms. Organizations should follow existing regulations and guidelines for risk criteria, tolerance, and response established by organizational, domain, discipline, sector, or professional requirements. Some sectors or industries may have established definitions of harm or established documentation, reporting, and disclosure requirements. Within sectors, risk management may depend on existing guidelines for specific applications and use case settings. Where established guidelines do not exist, organizations should define reasonable risk tolerance. Once tolerance is defined, this AI RMF can be used to manage risks and to document risk management processes. 1.2.3 Risk Prioritization Attempting to eliminate negative risk entirely can be counterproductive in practice because not all incidents and failures can be eliminated. Unrealistic expectations about risk may lead organizations to allocate resources in a manner that makes risk triage inefficient or impractical or wastes scarce resources. A risk management culture can help organizations recognize that not all AI risks are the same, and resources can be allocated purposefully. Actionable risk management efforts lay out clear guidelines for assessing trustworthiness of each AI system an organization develops or deploys. Policies and resources should be prioritized based on the assessed risk level and potential impact of an AI system. The extent to which an AI system may be customized or tailored to the specific context of use by the Al deployer can be a contributing factor. Page 7 25 NIST Al 100-1 AI RMF 1.0 When applying the AI RMF, risks which the organization determines to be highest for the AI systems within a given context of use call for the most urgent prioritization and most thorough risk management process. In cases where an AI system presents unacceptable negative risk levels — such as where significant negative impacts are imminent, severe harms are actually occurring, or catastrophic risks are present — development and deployment should cease in a safe manner until risks can be sufficiently managed. If an AI system's development, deployment, and use cases are found to be low -risk in a specific context, that may suggest potentially lower prioritization. Risk prioritization may differ between AI systems that are designed or deployed to directly interact with humans as compared to AI systems that are not. Higher initial prioritization may be called for in settings where the AI system is trained on large datasets comprised of sensitive or protected data such as personally identifiable information, or where the outputs of the AI systems have direct or indirect impact on humans. AI systems designed to interact only with computational systems and trained on non -sensitive datasets (for example, data collected from the physical environment) may call for lower initial prioritization. Nonethe- less, regularly assessing and prioritizing risk based on context remains important because non -human -facing AI systems can have downstream safety or social implications. Residual risk — defined as risk remaining after risk treatment (Source: ISO GUIDE 73) — directly impacts end users or affected individuals and communities. Documenting residual risks will call for the system provider to fully consider the risks of deploying the AI product and will inform end users about potential negative impacts of interacting with the system. 1.2.4 Organizational Integration and Management of Risk AI risks should not be considered in isolation. Different AI actors have different responsi- bilities and awareness depending on their roles in the lifecycle. For example, organizations developing an AI system often will not have information about how the system may be used. AI risk management should be integrated and incorporated into broader enterprise risk management strategies and processes. Treating AI risks along with other critical risks, such as cybersecurity and privacy, will yield a more integrated outcome and organizational efficiencies. The AI RMF may be utilized along with related guidance and frameworks for managing AI system risks or broader enterprise risks. Some risks related to AI systems are common across other types of software development and deployment. Examples of overlapping risks include: privacy concerns related to the use of underlying data to train AI systems; the en- ergy and environmental implications associated with resource -heavy computing demands; security concerns related to the confidentiality, integrity, and availability of the system and its training and output data; and general security of the underlying software and hardware for AI systems. Page 8 26 NIST Al 100-1 AI RMF 1.0 Organizations need to establish and maintain the appropriate accountability mechanisms, roles and responsibilities, culture, and incentive structures for risk management to be ef- fective. Use of the AI RMF alone will not lead to these changes or provide the appropriate incentives. Effective risk management is realized through organizational commitment at senior levels and may require cultural change within an organization or industry. In addi- tion, small to medium-sized organizations managing AI risks or implementing the AI RMF may face different challenges than large organizations, depending on their capabilities and resources. 2. Audience Identifying and managing AI risks and potential impacts — both positive and negative — re- quires a broad set of perspectives and actors across the AI lifecycle. Ideally, AI actors will represent a diversity of experience, expertise, and backgrounds and comprise demograph- ically and disciplinarily diverse teams. The AI RMF is intended to be used by AI actors across the AI lifecycle and dimensions. The OECD has developed a framework for classifying AI lifecycle activities according to five key socio-technical dimensions, each with properties relevant for AI policy and gover- nance, including risk management [OECD (2022) OECD Framework for the Classification of AI systems — OECD Digital Economy Papers]. Figure 2 shows these dimensions, slightly modified by NIST for purposes of this framework. The NIST modification high- lights the importance of test, evaluation, verification, and validation (TEVV) processes throughout an AI lifecycle and generalizes the operational context of an AI system. AI dimensions displayed in Figure 2 are the Application Context, Data and Input, AI Model, and Task and Output. AI actors involved in these dimensions who perform or manage the design, development, deployment, evaluation, and use of AI systems and drive AI risk management efforts are the primary AI RMF audience. Representative AI actors across the lifecycle dimensions are listed in Figure 3 and described in detail in Appendix A. Within the AI RMF, all AI actors work together to manage risks and achieve the goals of trustworthy and responsible AI. AI actors with TEVV-specific expertise are integrated throughout the AI lifecycle and are especially likely to benefit from the Framework. Performed regularly, TEVV tasks can provide insights relative to technical, societal, legal, and ethical standards or norms, and can assist with anticipating impacts and assessing and tracking emergent risks. As a regular process within an AI lifecycle, TEVV allows for both mid -course remediation and post -hoc risk management. The People & Planet dimension at the center of Figure 2 represents human rights and the broader well-being of society and the planet. The AI actors in this dimension comprise a separate AI RMF audience who informs the primary audience. These AI actors may in- clude trade associations, standards developing organizations, researchers, advocacy groups, Page 9 27 MST Al 100-1 AI RMF 1.0 X,e Verify 'aJa\`aa AL_ Fig. 2. Lifecycle and Key Dimensions of an AI System. Modified from OECD (2022) OECD Framework for the Classification of AI systems — OECD Digital Economy Papers. The two inner circles show Al systems' key dimensions and the outer circle shows Al lifecycle stages. Ideally, risk management efforts start with the Plan and Design function in the application context and are performed throughout the AI system lifecycle. See Figure 3 for representative AI actors. environmental groups, civil society organizations, end users, and potentially impacted in- dividuals and communities. These actors can: • assist in providing context and understanding potential and actual impacts; • be a source of formal or quasi -formal norms and guidance for AI risk management; • designate boundaries for Al operation (technical, societal, legal, and ethical); and • promote discussion of the tradeoffs needed to balance societal values and priorities related to civil liberties and rights, equity, the environment and the planet, and the economy. Successful risk management depends upon a sense of collective responsibility among AI actors shown in Figure 3. The AI RMF functions, described in Section 5, require diverse perspectives, disciplines, professions, and experiences. Diverse teams contribute to more open sharing of ideas and assumptions about the purposes and functions of technology — making these implicit aspects more explicit. This broader collective perspective creates opportunities for surfacing problems and identifying existing and emergent risks. Page 10 28 NIST AI 100-1 Al RMF 1.0 E Y V c m C a! m C T O h U UO m u K D n 5 a!2Eoa �uvEa�m°°c'ur-w E > N N 0— N O •E 'u O. O. •E > Q E C a! O � .'C 'a J O. N N is ccEIaa))Eolyo2ool� mEm wm.- ooiEomrn`a!o E • 'tJ J C .00 V! T o L O .� Cw c m a! r m = , m 0 m E >'? o ° C o o+ " acv E o` ' IQ 6 �a N N ate+ U C • O U C m y C E U m m _ m 0 Li • o > V m m a! C 0 O. N C 0 E U o u E v p m J a! C N • E U V m m Y F va, 'C'C > > > 'O a! 'CO v E .3 a) m � c a! - —" C O Y O m C O v m u y c _ > m >a w O r E C 00 L v 0,0 u m E C ac E o. a On °o o u; c y m O E O a! 'C22.2 I v .n u ar alycca`!a`lauuJi`oo>> V vlv0 C C X X J X O O w a 02 vaa�du .nrnr m E ci)_ -0 c 0 m i W N m 0 a 2o O yL.. '2 2m L al ' 2 E V m y .2x.2 ar- a,a u c O ON • va i_. N L U O U O c7 m E u a o 'C ac! a ax! m 2a ax! c c u a C J c i yC C _ O Y c 'Ea as ao!-omL�ma` °E6oaEoouau a ^o m 1ps C ' V c m ' O C N C C o ax E O 7 d > }' N vOi a! Qiii QO. mc' EN m'.' ` 011=11111 d c�°J>E a�uo> a�`m >cXEwmX'>ommx�o> v! m a,. r E w a! `o E a,. u w suoisuaun4 aEeas AA31 saiainipy /aN OIDADa;i1 sJOPV en! eauesaadaa Page 11 29 MST Al 100-1 AI RMF 1.0 3. AI Risks and Trustworthiness For AI systems to be trustworthy, they often need to be responsive to a multiplicity of cri- teria that are of value to interested parties. Approaches which enhance AI trustworthiness can reduce negative AI risks. This Framework articulates the following characteristics of trustworthy AI and offers guidance for addressing them. Characteristics of trustworthy AI systems include: valid and reliable, safe, secure and resilient, accountable and trans- parent, explainable and interpretable, privacy -enhanced, and fair with harmful bias managed. Creating trustworthy AI requires balancing each of these characteristics based on the AI system's context of use. While all characteristics are socio-technical system at- tributes, accountability and transparency also relate to the processes and activities internal to an AI system and its external setting. Neglecting these characteristics can increase the probability and magnitude of negative consequences. Fig. 4. Characteristics of trustworthy AI systems. Valid & Reliable is a necessary condition of trustworthiness and is shown as the base for other trustworthiness characteristics. Accountable & Transparent is shown as a vertical box because it relates to all other characteristics. Trustworthiness characteristics (shown in Figure 4) are inextricably tied to social and orga- nizational behavior, the datasets used by AI systems, selection of AI models and algorithms and the decisions made by those who build them, and the interactions with the humans who provide insight from and oversight of such systems. Human judgment should be employed when deciding on the specific metrics related to AI trustworthiness characteristics and the precise threshold values for those metrics. Addressing AI trustworthiness characteristics individually will not ensure AI system trust- worthiness; tradeoffs are usually involved, rarely do all characteristics apply in every set- ting, and some will be more or less important in any given situation. Ultimately, trustwor- thiness is a social concept that ranges across a spectrum and is only as strong as its weakest characteristics. When managing AI risks, organizations can face difficult decisions in balancing these char- acteristics. For example, in certain scenarios tradeoffs may emerge between optimizing for interpretability and achieving privacy. In other cases, organizations might face a tradeoff between predictive accuracy and interpretability. Or, under certain conditions such as data sparsity, privacy -enhancing techniques can result in a loss in accuracy, affecting decisions Page 12 30 NIST Al 100-1 AI RMF 1.0 about fairness and other values in certain domains. Dealing with tradeoffs requires tak- ing into account the decision -making context. These analyses can highlight the existence and extent of tradeoffs between different measures, but they do not answer questions about how to navigate the tradeoff. Those depend on the values at play in the relevant context and should be resolved in a manner that is both transparent and appropriately justifiable. There are multiple approaches for enhancing contextual awareness in the AI lifecycle. For example, subject matter experts can assist in the evaluation of TEVV findings and work with product and deployment teams to align TEVV parameters to requirements and de- ployment conditions. When properly resourced, increasing the breadth and diversity of input from interested parties and relevant AI actors throughout the AI lifecycle can en- hance opportunities for informing contextually sensitive evaluations, and for identifying AI system benefits and positive impacts. These practices can increase the likelihood that risks arising in social contexts are managed appropriately. Understanding and treatment of trustworthiness characteristics depends on an AI actor's particular role within the AI lifecycle. For any given AI system, an AI designer or developer may have a different perception of the characteristics than the deployer. Trustworthiness characteristics explained in this document influence each other. Highly secure but unfair systems, accurate but opaque and uninterpretable systems, and inaccurate but secure, privacy -enhanced, and transparent systems are all unde- sirable. A comprehensive approach to risk management calls for balancing tradeoffs among the trustworthiness characteristics. It is the joint responsibility of all AI ac- tors to determine whether AI technology is an appropriate or necessary tool for a given context or purpose, and how to use it responsibly. The decision to commission or deploy an AI system should be based on a contextual assessment of trustworthi- ness characteristics and the relative risks, impacts, costs, and benefits, and informed by a broad set of interested parties. 3.1 Valid and Reliable Validation is the "confirmation, through the provision of objective evidence, that the re- quirements for a specific intended use or application have been fulfilled" (Source: Iso 9000:2015). Deployment of AI systems which are inaccurate, unreliable, or poorly gener- alized to data and settings beyond their training creates and increases negative AI risks and reduces trustworthiness. Reliability is defined in the same standard as the "ability of an item to perform as required, without failure, for a given time interval, under given conditions" (Source: ISO/IEC TS 5723:2022). Reliability is a goal for overall correctness of AI system operation under the conditions of expected use and over a given period of time, including the entire lifetime of the system. Page 13 31 NIST Al 100-1 AI RMF 1.0 Accuracy and robustness contribute to the validity and trustworthiness of AI systems, and can be in tension with one another in AI systems. Accuracy is defined by ISO/IEC Ts 5723:2022 as "closeness of results of observations, computations, or estimates to the true values or the values accepted as being true." Mea- sures of accuracy should consider computational -centric measures (e.g., false positive and false negative rates), human -AI teaming, and demonstrate external validity (generalizable beyond the training conditions). Accuracy measurements should always be paired with clearly defined and realistic test sets — that are representative of conditions of expected use — and details about test methodology; these should be included in associated documen- tation. Accuracy measurements may include disaggregation of results for different data segments. Robustness or generalizability is defined as the "ability of a system to maintain its level of performance under a variety of circumstances" (Source: ISO/IEC TS 5723:2022). Ro- bustness is a goal for appropriate system functionality in a broad set of conditions and circumstances, including uses of AI systems not initially anticipated. Robustness requires not only that the system perform exactly as it does under expected uses, but also that it should perform in ways that minimize potential harms to people if it is operating in an unexpected setting. Validity and reliability for deployed AI systems are often assessed by ongoing testing or monitoring that confirms a system is performing as intended. Measurement of validity, accuracy, robustness, and reliability contribute to trustworthiness and should take into con- sideration that certain types of failures can cause greater harm. AI risk management efforts should prioritize the minimization of potential negative impacts, and may need to include human intervention in cases where the AI system cannot detect or correct errors. 3.2 Safe AI systems should "not under defined conditions, lead to a state in which human life, health, property, or the environment is endangered" (Source: ISO/IEC Ts 5723:2022). Safe operation of AI systems is improved through: • responsible design, development, and deployment practices; • clear information to deployers on responsible use of the system; • responsible decision -making by deployers and end users; and • explanations and documentation of risks based on empirical evidence of incidents. Different types of safety risks may require tailored AI risk management approaches based on context and the severity of potential risks presented. Safety risks that pose a potential risk of serious injury or death call for the most urgent prioritization and most thorough risk management process. Page 14 32 NIST Al 100-1 AI RMF 1.0 Employing safety considerations during the lifecycle and starting as early as possible with planning and design can prevent failures or conditions that can render a system dangerous. Other practical approaches for AI safety often relate to rigorous simulation and in -domain testing, real-time monitoring, and the ability to shut down, modify, or have human inter- vention into systems that deviate from intended or expected functionality. AI safety risk management approaches should take cues from efforts and guidelines for safety in fields such as transportation and healthcare, and align with existing sector- or application -specific guidelines or standards. 3.3 Secure and Resilient AI systems, as well as the ecosystems in which they are deployed, may be said to be re- silient if they can withstand unexpected adverse events or unexpected changes in their envi- ronment or use — or if they can maintain their functions and structure in the face of internal and external change and degrade safely and gracefully when this is necessary (Adapted from: ISO/IEC Ts 5723:2022). Common security concerns relate to adversarial examples, data poisoning, and the exfiltration of models, training data, or other intellectual property through AI system endpoints. AI systems that can maintain confidentiality, integrity, and availability through protection mechanisms that prevent unauthorized access and use may be said to be secure. Guidelines in the NIST Cybersecurity Framework and Risk Manage- ment Framework are among those which are applicable here. Security and resilience are related but distinct characteristics. While resilience is the abil- ity to return to normal function after an unexpected adverse event, security includes re- silience but also encompasses protocols to avoid, protect against, respond to, or recover from attacks. Resilience relates to robustness and goes beyond the provenance of the data to encompass unexpected or adversarial use (or abuse or misuse) of the model or data. 3.4 Accountable and Transparent Trustworthy AI depends upon accountability. Accountability presupposes transparency. Transparency reflects the extent to which information about an AI system and its outputs is available to individuals interacting with such a system — regardless of whether they are even aware that they are doing so. Meaningful transparency provides access to appropriate levels of information based on the stage of the AI lifecycle and tailored to the role or knowledge of AI actors or individuals interacting with or using the AI system. By promoting higher levels of understanding, transparency increases confidence in the AI system. This characteristic's scope spans from design decisions and training data to model train- ing, the structure of the model, its intended use cases, and how and when deployment, post -deployment, or end user decisions were made and by whom. Transparency is often necessary for actionable redress related to AI system outputs that are incorrect or otherwise lead to negative impacts. Transparency should consider human -AI interaction: for exam - Page 15 33 MST Al 100-1 AI RMF 1.0 pie, how a human operator or user is notified when a potential or actual adverse outcome caused by an AI system is detected. A transparent system is not necessarily an accurate, privacy -enhanced, secure, or fair system. However, it is difficult to determine whether an opaque system possesses such characteristics, and to do so over time as complex systems evolve. The role of AI actors should be considered when seeking accountability for the outcomes of AI systems. The relationship between risk and accountability associated with AI and tech- nological systems more broadly differs across cultural, legal, sectoral, and societal contexts. When consequences are severe, such as when life and liberty are at stake, AI developers and deployers should consider proportionally and proactively adjusting their transparency and accountability practices. Maintaining organizational practices and governing structures for harm reduction, like risk management, can help lead to more accountable systems. Measures to enhance transparency and accountability should also consider the impact of these efforts on the implementing entity, including the level of necessary resources and the need to safeguard proprietary information. Maintaining the provenance of training data and supporting attribution of the AI system's decisions to subsets of training data can assist with both transparency and accountability. Training data may also be subject to copyright and should follow applicable intellectual property rights laws. As transparency tools for AI systems and related documentation continue to evolve, devel- opers of AI systems are encouraged to test different types of transparency tools in cooper- ation with AI deployers to ensure that AI systems are used as intended. 3.5 Explainable and Interpretable Explainability refers to a representation of the mechanisms underlying AI systems' oper- ation, whereas interpretability refers to the meaning of AI systems' output in the context of their designed functional purposes. Together, explainability and interpretability assist those operating or overseeing an AI system, as well as users of an AI system, to gain deeper insights into the functionality and trustworthiness of the system, including its out- puts. The underlying assumption is that perceptions of negative risk stem from a lack of ability to make sense of, or contextualize, system output appropriately. Explainable and interpretable AI systems offer information that will help end users understand the purposes and potential impact of an AI system. Risk from lack of explainability may be managed by describing how AI systems function, with descriptions tailored to individual differences such as the user's role, knowledge, and skill level. Explainable systems can be debugged and monitored more easily, and they lend themselves to more thorough documentation, audit, and governance. Page 16 34 NIST Al 100-1 AI RMF 1.0 Risks to interpretability often can be addressed by communicating a description of why an AI system made a particular prediction or recommendation. (See "Four Principles of Explainable Artificial Intelligence" and "Psychological Foundations of Explainability and Interpretability in Artificial Intelligence" found here.) Transparency, explainability, and interpretability are distinct characteristics that support each other. Transparency can answer the question of "what happened" in the system. Ex- plainability can answer the question of "how" a decision was made in the system. Inter- pretability can answer the question of "why" a decision was made by the system and its meaning or context to the user. 3.6 Privacy -Enhanced Privacy refers generally to the norms and practices that help to safeguard human autonomy, identity, and dignity. These norms and practices typically address freedom from intrusion, limiting observation, or individuals' agency to consent to disclosure or control of facets of their identities (e.g., body, data, reputation). (See The NIST Privacy Framework: A Tool for Improving Privacy through Enterprise Risk Management.) Privacy values such as anonymity, confidentiality, and control generally should guide choices for AI system design, development, and deployment. Privacy -related risks may influence security, bias, and transparency and come with tradeoffs with these other characteristics. Like safety and security, specific technical features of an AI system may promote or reduce privacy. AI systems can also present new risks to privacy by allowing inference to identify individuals or previously private information about individuals. Privacy -enhancing technologies ("PETs") for AI, as well as data minimizing methods such as de -identification and aggregation for certain model outputs, can support design for privacy -enhanced AI systems. Under certain conditions such as data sparsity, privacy - enhancing techniques can result in a loss in accuracy, affecting decisions about fairness and other values in certain domains. 3.7 Fair — with Harmful Bias Managed Fairness in AI includes concerns for equality and equity by addressing issues such as harm- ful bias and discrimination. Standards of fairness can be complex and difficult to define be- cause perceptions of fairness differ among cultures and may shift depending on application. Organizations' risk management efforts will be enhanced by recognizing and considering these differences. Systems in which harmful biases are mitigated are not necessarily fair. For example, systems in which predictions are somewhat balanced across demographic groups may still be inaccessible to individuals with disabilities or affected by the digital divide or may exacerbate existing disparities or systemic biases. Page 17 35 NIST Al 100-1 AI RMF 1.0 Bias is broader than demographic balance and data representativeness. NIST has identified three major categories of Al bias to be considered and managed: systemic, computational and statistical, and human -cognitive. Each of these can occur in the absence of prejudice, partiality, or discriminatory intent. Systemic bias can be present in AI datasets, the orga- nizational norms, practices, and processes across the Al lifecycle, and the broader society that uses AI systems. Computational and statistical biases can be present in AI datasets and algorithmic processes, and often stem from systematic errors due to non -representative samples. Human -cognitive biases relate to how an individual or group perceives AI sys- tem information to make a decision or fill in missing information, or how humans think about purposes and functions of an AI system. Human -cognitive biases are omnipresent in decision -making processes across the AI lifecycle and system use, including the design, implementation, operation, and maintenance of AI. Bias exists in many forms and can become ingrained in the automated systems that help make decisions about our lives. While bias is not always a negative phenomenon, AI sys- tems can potentially increase the speed and scale of biases and perpetuate and amplify harms to individuals, groups, communities, organizations, and society. Bias is tightly asso- ciated with the concepts of transparency as well as fairness in society. (For more informa- tion about bias, including the three categories, see NIST Special Publication 1270, Towards a Standard for Identifying and Managing Bias in Artificial Intelligence.) Page 18 36 NIST AI 100-1 Al RMF 1.0 4. Effectiveness of the AI RMF Evaluations of AI RMF effectiveness — including ways to measure bottom -line improve- ments in the trustworthiness of AI systems — will be part of future NIST activities, in conjunction with the AI community. Organizations and other users of the Framework are encouraged to periodically evaluate whether the AI RMF has improved their ability to manage AI risks, including but not lim- ited to their policies, processes, practices, implementation plans, indicators, measurements, and expected outcomes. NIST intends to work collaboratively with others to develop met- rics, methodologies, and goals for evaluating the AI RMF's effectiveness, and to broadly share results and supporting information. Framework users are expected to benefit from: • enhanced processes for governing, mapping, measuring, and managing AI risk, and clearly documenting outcomes; • improved awareness of the relationships and tradeoffs among trustworthiness char- acteristics, socio-technical approaches, and AI risks; • explicit processes for making go/no-go system commissioning and deployment deci- sions; • established policies, processes, practices, and procedures for improving organiza- tional accountability efforts related to AI system risks; • enhanced organizational culture which prioritizes the identification and management of AI system risks and potential impacts to individuals, communities, organizations, and society; • better information sharing within and across organizations about risks, decision - making processes, responsibilities, common pitfalls, TEVV practices, and approaches for continuous improvement; • greater contextual knowledge for increased awareness of downstream risks; • strengthened engagement with interested parties and relevant AI actors; and • augmented capacity for TEVV of AI systems and associated risks. Page 19 37 NIST AI 100-1 AI RMF 1.0 Part 2: Core and Profiles 5. AI RMF Core The AI RMF Core provides outcomes and actions that enable dialogue, understanding, and activities to manage AI risks and responsibly develop trustworthy AI systems. As illus- trated in Figure 5, the Core is composed of four functions: GOVERN, MAP, MEASURE, and MANAGE. Each of these high-level functions is broken down into categories and sub- categories. Categories and subcategories are subdivided into specific actions and outcomes. Actions do not constitute a checklist, nor are they necessarily an ordered set of steps. Map Context is recognized and risks related to context are identified Measure ed risks sessed, led, or -ked Fig. 5. Functions organize AI risk management activities at their highest level to govern, map, measure, and manage AI risks. Governance is designed to be a cross -cutting function to inform and be infused throughout the other three functions. Risk management should be continuous, timely, and performed throughout the AI system lifecycle dimensions. AI RMF Core functions should be carried out in a way that reflects diverse and multidisciplinary perspectives, potentially including the views of AI actors out- side the organization. Having a diverse team contributes to more open sharing of ideas and assumptions about purposes and functions of the technology being designed, developed, Page 20 38 NIST AI 100-1 Al RMF 1.0 deployed, or evaluated — which can create opportunities to surface problems and identify existing and emergent risks. An online companion resource to the AI RMF, the NIST AI RMF Playbook, is available to help organizations navigate the AI RMF and achieve its outcomes through suggested tactical actions they can apply within their own contexts. Like the Al RMF, the Playbook is voluntary and organizations can utilize the suggestions according to their needs and interests. Playbook users can create tailored guidance selected from suggested material for their own use and contribute their suggestions for sharing with the broader community. Along with the AI RMF, the Playbook is part of the NIST Trustworthy and Responsible AI Resource Center. Framework users may apply these functions as best suits their needs for managing AI risks based on their resources and capabilities. Some organizations may choose to select from among the categories and subcategories; others may choose and have the capacity to apply all categories and subcategories. Assuming a governance struc- ture is in place, functions may be performed in any order across the Al lifecycle as deemed to add value by a user of the framework. After instituting the outcomes in GOVERN, most users of the AI RMF would start with the MAP function and con- tinue to MEASURE or MANAGE. However users integrate the functions, the process should be iterative, with cross-referencing between functions as necessary. Simi- larly, there are categories and subcategories with elements that apply to multiple functions, or that logically should take place before certain subcategory decisions. 5.1 Govern The GOVERN function: • cultivates and implements a culture of risk management within organizations design- ing, developing, deploying, evaluating, or acquiring AI systems; • outlines processes, documents, and organizational schemes that anticipate, identify, and manage the risks a system can pose, including to users and others across society — and procedures to achieve those outcomes; • incorporates processes to assess potential impacts; • provides a structure by which AI risk management functions can align with organi- zational principles, policies, and strategic priorities; • connects technical aspects of Al system design and development to organizational values and principles, and enables organizational practices and competencies for the individuals involved in acquiring, training, deploying, and monitoring such systems; and • addresses full product lifecycle and associated processes, including legal and other issues concerning use of third -party software or hardware systems and data. Page 21 39 NIST AI 100-1 Al RMF 1.0 GOVERN is a cross -cutting function that is infused throughout AI risk management and enables the other functions of the process. Aspects of GOVERN, especially those related to compliance or evaluation, should be integrated into each of the other functions. Attention to governance is a continual and intrinsic requirement for effective AI risk management over an AI system's lifespan and the organization's hierarchy. Strong governance can drive and enhance internal practices and norms to facilitate orga- nizational risk culture. Governing authorities can determine the overarching policies that direct an organization's mission, goals, values, culture, and risk tolerance. Senior leader- ship sets the tone for risk management within an organization, and with it, organizational culture. Management aligns the technical aspects of AI risk management to policies and operations. Documentation can enhance transparency, improve human review processes, and bolster accountability in AI system teams. After putting in place the structures, systems, processes, and teams described in the GOV- ERN function, organizations should benefit from a purpose -driven culture focused on risk understanding and management. It is incumbent on Framework users to continue to ex- ecute the GOVERN function as knowledge, cultures, and needs or expectations from AI actors evolve over time. Practices related to governing AI risks are described in the NIST AI RMF Playbook. Table 1 lists the GOVERN function's categories and subcategories. Table 1: Categories and subcategories for the GOVERN function. Categories Subcategories GOVERN 1: GOVERN 1.1: Legal and regulatory requirements involving AI Policies, processes, are understood, managed, and documented. procedures, and GOVERN 1.2: The characteristics of trustworthy AI are inte- practices across the grated into organizational policies, processes, procedures, and organization related practices. to the mapping, measuring, and GOVERN 1.3: Processes, procedures, and practices are in place managing of AI to determine the needed level of risk management activities based risks are in place, on the organization's risk tolerance. transparent, and GOVERN 1.4: The risk management process and its outcomes are implemented established through transparent policies, procedures, and other effectively, controls based on organizational risk priorities. Continued on next page Page 22 40 NIST AI 100-1 Al RMF 1.0 Table 1: Categories and subcategories for the GOVERN function. (Continued) Categories Subcategories GOVERN 1.5: Ongoing monitoring and periodic review of the risk management process and its outcomes are planned and or- ganizational roles and responsibilities clearly defined, including determining the frequency of periodic review. GOVERN 1.6: Mechanisms are in place to inventory AI systems and are resourced according to organizational risk priorities. GOVERN 1.7: Processes and procedures are in place for decom- missioning and phasing out AI systems safely and in a man- ner that does not increase risks or decrease the organization's trustworthiness. GOVERN 2: GOVERN 2.1: Roles and responsibilities and lines of communi- Accountability cation related to mapping, measuring, and managing AI risks are structures are in documented and are clear to individuals and teams throughout place so that the the organization. appropriate teams GOVERN 2.2: The organization's personnel and partners receive and individuals are AI risk management training to enable them to perform their du - empowered, ties and responsibilities consistent with related policies, proce- responsible, and dures, and agreements. trained for mapping, and measuring,managing GOVERN 2.3: Executive leadership of the organization takes re- Al risks. sponsibility for decisions about risks associated with AI system development and deployment. GOVERN 3: GOVERN 3.1: Decision -making related to mapping, measuring, Workforce diversity, and managing AI risks throughout the lifecycle is informed by a equity, inclusion, diverse team (e.g., diversity of demographics, disciplines, expe- and accessibility rience, expertise, and backgrounds). processes are prioritized in the GOVERN 3.2: Policies and procedures are in place to define and mapping, differentiate roles and responsibilities for human -AI configura- measuring, and tions and oversight of AI systems. managing of AI risks throughout the lifecycle. GOVERN 4: GOVERN 4.1: Organizational policies and practices are in place Organizational to foster a critical thinking and safety -first mindset in the design, teams are committed development, deployment, and uses of AI systems to minimize to a culture potential negative impacts. Continued on next page Page 23 41 MST Al 100-1 Al RMF 1.0 Table 1: Categories and subcategories for the GOVERN function. (Continued) Categories Subcategories that considers and GOVERN 4.2: Organizational teams document the risks and po- communicates AI tential impacts of the AI technology they design, develop, deploy, risk. evaluate, and use, and they communicate about the impacts more broadly. GOVERN 4.3: Organizational practices are in place to enable AI testing, identification of incidents, and information sharing. GOVERN 5: GOVERN 5.1: Organizational policies and practices are in place Processes are in to collect, consider, prioritize, and integrate feedback from those place for robust external to the team that developed or deployed the AI system engagement with regarding the potential individual and societal impacts related to relevant AI actors. AI risks. GOVERN 5.2: Mechanisms are established to enable the team that developed or deployed AI systems to regularly incorporate adjudicated feedback from relevant AI actors into system design and implementation. GOVERN 6: Policies GOVERN 6.1: Policies and procedures are in place that address and procedures are AI risks associated with third -party entities, including risks of in - in place to address fringement of a third -party's intellectual property or other rights. AI risks and benefits GOVERN 6.2: Contingency processes are in place to handle arising from failures or incidents in third -party data or AI systems deemed to third -party software be high -risk. and data and other supply chain issues. 5.2 Map The MAP function establishes the context to frame risks related to an AI system. The AI lifecycle consists of many interdependent activities involving a diverse set of actors (See Figure 3). In practice, AI actors in charge of one part of the process often do not have full visibility or control over other parts and their associated contexts. The interdependencies between these activities, and among the relevant AI actors, can make it difficult to reliably anticipate impacts of AI systems. For example, early decisions in identifying purposes and objectives of an AI system can alter its behavior and capabilities, and the dynamics of de- ployment setting (such as end users or impacted individuals) can shape the impacts of AI system decisions. As a result, the best intentions within one dimension of the AI lifecycle can be undermined via interactions with decisions and conditions in other, later activities. Page 24 42 NIST Al 100-1 AI RMF 1.0 This complexity and varying levels of visibility can introduce uncertainty into risk man- agement practices. Anticipating, assessing, and otherwise addressing potential sources of negative risk can mitigate this uncertainty and enhance the integrity of the decision process. The information gathered while carrying out the MAP function enables negative risk pre- vention and informs decisions for processes such as model management, as well as an initial decision about appropriateness or the need for an AI solution. Outcomes in the MAP function are the basis for the MEASURE and MANAGE functions. Without contex- tual knowledge, and awareness of risks within the identified contexts, risk management is difficult to perform. The MAP function is intended to enhance an organization's ability to identify risks and broader contributing factors. Implementation of this function is enhanced by incorporating perspectives from a diverse internal team and engagement with those external to the team that developed or deployed the AI system. Engagement with external collaborators, end users, potentially impacted communities, and others may vary based on the risk level of a particular AI system, the makeup of the internal team, and organizational policies. Gathering such broad perspec- tives can help organizations proactively prevent negative risks and develop more trustwor- thy AI systems by: • improving their capacity for understanding contexts; • checking their assumptions about context of use; • enabling recognition of when systems are not functional within or out of their in- tended context; • identifying positive and beneficial uses of their existing AI systems; • improving understanding of limitations in AI and ML processes; • identifying constraints in real -world applications that may lead to negative impacts; • identifying known and foreseeable negative impacts related to intended use of AI systems; and • anticipating risks of the use of AI systems beyond intended use. After completing the MAP function, Framework users should have sufficient contextual knowledge about AI system impacts to inform an initial go/no-go decision about whether to design, develop, or deploy an AI system. If a decision is made to proceed, organizations should utilize the MEASURE and MANAGE functions along with policies and procedures put into place in the GOVERN function to assist in AI risk management efforts. It is incum- bent on Framework users to continue applying the MAP function to AI systems as context, capabilities, risks, benefits, and potential impacts evolve over time. Practices related to mapping AI risks are described in the NIST AI RMF Playbook. Table 2 lists the MAP function's categories and subcategories. Page 25 43 MST Al 100-1 Al RMF 1.0 Table 2: Categories and subcategories for the MAP function. Categories Subcategories MAP 1: Context is MAP 1.1: Intended purposes, potentially beneficial uses, context - established and specific laws, norms and expectations, and prospective settings in understood. which the AI system will be deployed are understood and docu- mented. Considerations include: the specific set or types of users along with their expectations; potential positive and negative im- pacts of system uses to individuals, communities, organizations, society, and the planet; assumptions and related limitations about AI system purposes, uses, and risks across the development or product AI lifecycle; and related TEVV and system metrics. MAP 1.2: Interdisciplinary AI actors, competencies, skills, and capacities for establishing context reflect demographic diversity and broad domain and user experience expertise, and their par- ticipation is documented. Opportunities for interdisciplinary col- laboration are prioritized. MAP 1.3: The organization's mission and relevant goals for AI technology are understood and documented. MAP 1.4: The business value or context of business use has been clearly defined or — in the case of assessing existing AI systems — re-evaluated. MAP 1.5: Organizational risk tolerances are determined and documented. MAP 1.6: System requirements (e.g., "the system shall respect the privacy of its users") are elicited from and understood by rel- evant AI actors. Design decisions take socio-technical implica- tions into account to address AI risks. MAP 2: MAP 2.1: The specific tasks and methods used to implement the Categorization of tasks that the AI system will support are defined (e.g., classifiers, the AI system is generative models, recommenders). performed. MAP 2.2: Information about the AI system's knowledge limits and how system output may be utilized and overseen by humans is documented. Documentation provides sufficient information to assist relevant AI actors when making decisions and taking subsequent actions. Continued on next page Page 26 44 NIST AI 100-1 Al RMF 1.0 Table 2: Categories and subcategories for the MAP function. (Continued) Categories Subcategories MAP 2.3: Scientific integrity and TEVV considerations are iden- tified and documented, including those related to experimental design, data collection and selection (e.g., availability, repre- sentativeness, suitability), system trustworthiness, and construct validation. MAP 3: AI MAP 3.1: Potential benefits of intended AI system functionality capabilities, targeted and performance are examined and documented. usage, goals, and MAP 3.2: Potential costs, including non -monetary costs, which expected benefits result from expected or realized AI errors or system functionality and costs compared and trustworthiness — as connected to organizational risk toler- with appropriate ance — are examined and documented. benchmarks are understood. MAP 3.3: Targeted application scope is specified and docu- mented based on the system's capability, established context, and AI system categorization. MAP 3.4: Processes for operator and practitioner proficiency with AI system performance and trustworthiness — and relevant technical standards and certifications — are defined, assessed, and documented. MAP 3.5: Processes for human oversight are defined, assessed, and documented in accordance with organizational policies from the GOVERN function. MAP 4: Risks and MAP 4.1: Approaches for mapping AI technology and legal risks benefits are mapped of its components — including the use of third -party data or soft - for all components ware — are in place, followed, and documented, as are risks of in - of the AI system fringement of a third party's intellectual property or other rights. including third -party MAP 4.2: Internal risk controls for components of the AI sys- software and data. tem, including third -party AI technologies, are identified and documented. MAP 5: Impacts to MAP 5.1: Likelihood and magnitude of each identified impact individuals, groups, (both potentially beneficial and harmful) based on expected use, communities, past uses of AI systems in similar contexts, public incident re- organizations, and ports, feedback from those external to the team that developed society are or deployed the AI system, or other data are identified and characterized. documented. Continued on next page Page 27 45 NIST AI 100-1 AI RMF 1.0 Table 2: Categories and subcategories for the MAP function. (Continued) Categories Subcategories MAP 5.2: Practices and personnel for supporting regular en- gagement with relevant AI actors and integrating feedback about positive, negative, and unanticipated impacts are in place and documented. 5.3 Measure The MEASURE function employs quantitative, qualitative, or mixed -method tools, tech- niques, and methodologies to analyze, assess, benchmark, and monitor AI risk and related impacts. It uses knowledge relevant to AI risks identified in the MAP function and informs the MANAGE function. AI systems should be tested before their deployment and regu- larly while in operation. AI risk measurements include documenting aspects of systems' functionality and trustworthiness. Measuring AI risks includes tracking metrics for trustworthy characteristics, social impact, and human -AI configurations. Processes developed or adopted in the MEASURE function should include rigorous software testing and performance assessment methodologies with associated measures of uncertainty, comparisons to performance benchmarks, and formal- ized reporting and documentation of results. Processes for independent review can improve the effectiveness of testing and can mitigate internal biases and potential conflicts of inter- est. Where tradeoffs among the trustworthy characteristics arise, measurement provides a trace- able basis to inform management decisions. Options may include recalibration, impact mitigation, or removal of the system from design, development, production, or use, as well as a range of compensating, detective, deterrent, directive, and recovery controls. After completing the MEASURE function, objective, repeatable, or scalable test, evaluation, verification, and validation (TEVV) processes including metrics, methods, and methodolo- gies are in place, followed, and documented. Metrics and measurement methodologies should adhere to scientific, legal, and ethical norms and be carried out in an open and trans- parent process. New types of measurement, qualitative and quantitative, may need to be developed. The degree to which each measurement type provides unique and meaningful information to the assessment of AI risks should be considered. Framework users will en- hance their capacity to comprehensively evaluate system trustworthiness, identify and track existing and emergent risks, and verify efficacy of the metrics. Measurement outcomes will be utilized in the MANAGE function to assist risk monitoring and response efforts. It is in- cumbent on Framework users to continue applying the MEASURE function to AI systems as knowledge, methodologies, risks, and impacts evolve over time. Page 28 46 NIST Al 100-1 AI RMF 1.0 Practices related to measuring AI risks are described in the NIST AI RMF Playbook. Table 3 lists the MEASURE function's categories and subcategories. Table 3: Categories and subcategories for the MEASURE function. Categories Subcategories MEASURE 1: MEASURE 1.1: Approaches and metrics for measurement of AI Appropriate risks enumerated during the MAP function are selected for imple- methods and metrics mentation starting with the most significant AI risks. The risks are identified and or trustworthiness characteristics that will not — or cannot — be applied, measured are properly documented. MEASURE 1.2: Appropriateness of AI metrics and effectiveness of existing controls are regularly assessed and updated, including reports of errors and potential impacts on affected communities. MEASURE 1.3: Internal experts who did not serve as front-line developers for the system and/or independent assessors are in- volved in regular assessments and updates. Domain experts, users, AI actors external to the team that developed or deployed the AI system, and affected communities are consulted in support of assessments as necessary per organizational risk tolerance. MEASURE 2: AI MEASURE 2.1: Test sets, metrics, and details about the tools used systems are during TEVV are documented. evaluated for MEASURE 2.2: Evaluations involving human subjects meet ap- trustworthy plicable requirements (including human subject protection) and characteristics. are representative of the relevant population. MEASURE 2.3: AI system performance or assurance criteria are measured qualitatively or quantitatively and demonstrated for conditions similar to deployment setting(s). Measures are documented. MEASURE 2.4: The functionality and behavior of the AI sys- tem and its components — as identified in the MAP function — are monitored when in production. MEASURE 2.5: The AI system to be deployed is demonstrated to be valid and reliable. Limitations of the generalizability be- yond the conditions under which the technology was developed are documented. Continued on next page Page 29 47 NIST AI 100-1 Al RMF 1.0 Table 3: Categories and subcategories for the MEASURE function. (Continued) Categories Subcategories MEASURE 2.6: The AI system is evaluated regularly for safety risks — as identified in the MAP function. The AI system to be de- ployed is demonstrated to be safe, its residual negative risk does not exceed the risk tolerance, and it can fail safely, particularly if made to operate beyond its knowledge limits. Safety metrics re- flect system reliability and robustness, real-time monitoring, and response times for AI system failures. MEASURE 2.7: AI system security and resilience — as identified in the MAP function — are evaluated and documented. MEASURE 2.8: Risks associated with transparency and account- ability — as identified in the MAP function — are examined and documented. MEASURE 2.9: The AI model is explained, validated, and docu- mented, and AI system output is interpreted within its context — as identified in the MAP function — to inform responsible use and governance. MEASURE 2.10: Privacy risk of the AI system — as identified in the MAP function — is examined and documented. MEASURE 2.11: Fairness and bias — as identified in the MAP function — are evaluated and results are documented. MEASURE 2.12: Environmental impact and sustainability of AI model training and management activities — as identified in the MAP function — are assessed and documented. MEASURE 2.13: Effectiveness of the employed TEVV met- rics and processes in the MEASURE function are evaluated and documented. MEASURE 3: MEASURE 3.1: Approaches, personnel, and documentation are Mechanisms for in place to regularly identify and track existing, unanticipated, tracking identified and emergent AI risks based on factors such as intended and ac - AI risks over time tual performance in deployed contexts. are in place. MEASURE 3.2: Risk tracking approaches are considered for settings where AI risks are difficult to assess using currently available measurement techniques or where metrics are not yet available. Continued on next page Page 30 48 MST Al 100-1 AI RMF 1.0 Table 3: Categories and subcategories for the MEASURE function. (Continued) Categories Subcategories MEASURE 3.3: Feedback processes for end users and impacted communities to report problems and appeal system outcomes are established and integrated into AI system evaluation metrics. MEASURE 4: MEASURE 4.1: Measurement approaches for identifying AI risks Feedback about are connected to deployment context(s) and informed through efficacy of consultation with domain experts and other end users. Ap- measurement is proaches are documented. gathered and MEASURE 4.2: Measurement results regarding AI system trust - assessed. worthiness in deployment context(s) and across the AI lifecycle are informed by input from domain experts and relevant AI ac- tors to validate whether the system is performing consistently as intended. Results are documented. MEASURE 4.3: Measurable performance improvements or de- clines based on consultations with relevant AI actors, in- cluding affected communities, and field data about context - relevant risks and trustworthiness characteristics are identified and documented. 5.4 Manage The MANAGE function entails allocating risk resources to mapped and measured risks on a regular basis and as defined by the GOVERN function. Risk treatment comprises plans to respond to, recover from, and communicate about incidents or events. Contextual information gleaned from expert consultation and input from relevant AI actors — established in GOVERN and carried out in MAP - is utilized in this function to decrease the likelihood of system failures and negative impacts. Systematic documentation practices established in GOVERN and utilized in MAP and MEASURE bolster AI risk management efforts and increase transparency and accountability. Processes for assessing emergent risks are in place, along with mechanisms for continual improvement. After completing the MANAGE function, plans for prioritizing risk and regular monitoring and improvement will be in place. Framework users will have enhanced capacity to man- age the risks of deployed AI systems and to allocate risk management resources based on assessed and prioritized risks. It is incumbent on Framework users to continue to apply the MANAGE function to deployed AI systems as methods, contexts, risks, and needs or expectations from relevant AI actors evolve over time. Page 31 49 NIST Al 100-1 AI RMF 1.0 Practices related to managing AI risks are described in the NIST Al RMF Playbook. Table 4 lists the MANAGE function's categories and subcategories. Table 4: Categories and subcategories for the MANAGE function. Categories Subcategories MANAGE 1: Al MANAGE 1.1: A determination is made as to whether the AI risks based on system achieves its intended purposes and stated objectives and assessments and whether its development or deployment should proceed. other analytical MANAGE 1.2: Treatment of documented AI risks is prioritized output from the based on impact, likelihood, and available resources or methods. MAP and MEASURE functions are MANAGE 1.3: Responses to the AI risks deemed high priority, as prioritized, identified by the MAP function, are developed, planned, and doc- responded to, and umented. Risk response options can include mitigating, transfer - managed. ring, avoiding, or accepting. MANAGE 1.4: Negative residual risks (defined as the sum of all unmitigated risks) to both downstream acquirers of AI systems and end users are documented. MANAGE 2: MANAGE 2.1: Resources required to manage Al risks are taken Strategies to into account — along with viable non -AI alternative systems, ap- maximize AI proaches, or methods — to reduce the magnitude or likelihood of benefits and potential impacts. minimize negative MANAGE 2.2: Mechanisms are in place and applied to sustain impacts are planned, the value of deployed AI systems. prepared, implemented, MANAGE 2.3: Procedures are followed to respond to and recover documented, and from a previously unknown risk when it is identified. informed by input MANAGE 2.4: Mechanisms are in place and applied, and respon- from relevant Al sibilities are assigned and understood, to supersede, disengage, or actors. deactivate AI systems that demonstrate performance or outcomes inconsistent with intended use. MANAGE 3: AI MANAGE 3.1: Al risks and benefits from third -party resources risks and benefits are regularly monitored, and risk controls are applied and from third -party documented. entities are MANAGE 3.2: Pre -trained models which are used for develop - managed. ment are monitored as part of Al system regular monitoring and maintenance. Continued on next page Page 32 50 MST Al 100-1 AI RMF 1.0 Table 4: Categories and subcategories for the MANAGE function. (Continued) Categories Subcategories MANAGE 4: Risk MANAGE 4.1: Post -deployment AI system monitoring plans treatments, are implemented, including mechanisms for capturing and eval- including response uating input from users and other relevant AI actors, appeal and recovery, and and override, decommissioning, incident response, recovery, and communication change management. plans for the MANAGE 4.2: Measurable activities for continual improvements identified and are integrated into AI system updates and include regular engage - measured AI risks ment with interested parties, including relevant AI actors. are documented and monitored regularly. MANAGE 4.3: Incidents and errors are communicated to relevant AI actors, including affected communities. Processes for track- ing, responding to, and recovering from incidents and errors are followed and documented. 6. AI RMF Profiles AI RMF use -case profiles are implementations of the Al RMF functions, categories, and subcategories for a specific setting or application based on the requirements, risk tolerance, and resources of the Framework user: for example, an AI RMF hiring profile or an Al RMF fair housing profile. Profiles may illustrate and offer insights into how risk can be managed at various stages of the Al lifecycle or in specific sector, technology, or end -use applications. AI RMF profiles assist organizations in deciding how they might best manage AI risk that is well -aligned with their goals, considers legal/regulatory requirements and best practices, and reflects risk management priorities. AI RMF temporal profiles are descriptions of either the current state or the desired, target state of specific AI risk management activities within a given sector, industry, organization, or application context. An Al RMF Current Profile indicates how AI is currently being managed and the related risks in terms of current outcomes. A Target Profile indicates the outcomes needed to achieve the desired or target AI risk management goals. Comparing Current and Target Profiles likely reveals gaps to be addressed to meet AI risk management objectives. Action plans can be developed to address these gaps to fulfill outcomes in a given category or subcategory. Prioritization of gap mitigation is driven by the user's needs and risk management processes. This risk -based approach also enables Framework users to compare their approaches with other approaches and to gauge the resources needed (e.g., staffing, funding) to achieve Al risk management goals in a cost- effective, prioritized manner. Page 33 51 NIST AI 100-1 Al RMF 1.0 AI RMF cross-sectoral profiles cover risks of models or applications that can be used across use cases or sectors. Cross-sectoral profiles can also cover how to govern, map, measure, and manage risks for activities or business processes common across sectors such as the use of large language models, cloud -based services or acquisition. This Framework does not prescribe profile templates, allowing for flexibility in implemen- tation. Page 34 52 MST Al 100-1 AI RMF 1.0 Appendix A: Descriptions of AI Actor Tasks from Figures 2 and 3 AI Design tasks are performed during the Application Context and Data and Input phases of the AI lifecycle in Figure 2. AI Design actors create the concept and objectives of AI systems and are responsible for the planning, design, and data collection and processing tasks of the AI system so that the AI system is lawful and fit -for -purpose. Tasks include ar- ticulating and documenting the system's concept and objectives, underlying assumptions, context, and requirements; gathering and cleaning data; and documenting the metadata and characteristics of the dataset. AI actors in this category include data scientists, do- main experts, socio-cultural analysts, experts in the field of diversity, equity, inclusion, and accessibility, members of impacted communities, human factors experts (e.g., UX/UI design), governance experts, data engineers, data providers, system funders, product man- agers, third -party entities, evaluators, and legal and privacy governance. AI Development tasks are performed during the AI Model phase of the lifecycle in Figure 2. AI Development actors provide the initial infrastructure of AI systems and are responsi- ble for model building and interpretation tasks, which involve the creation, selection, cali- bration, training, and/or testing of models or algorithms. AI actors in this category include machine learning experts, data scientists, developers, third -party entities, legal and privacy governance experts, and experts in the socio-cultural and contextual factors associated with the deployment setting. AI Deployment tasks are performed during the Task and Output phase of the lifecycle in Figure 2. AI Deployment actors are responsible for contextual decisions relating to how the AI system is used to assure deployment of the system into production. Related tasks include piloting the system, checking compatibility with legacy systems, ensuring regu- latory compliance, managing organizational change, and evaluating user experience. AI actors in this category include system integrators, software developers, end users, oper- ators and practitioners, evaluators, and domain experts with expertise in human factors, socio-cultural analysis, and governance. Operation and Monitoring tasks are performed in the Application Context/Operate and Monitor phase of the lifecycle in Figure 2. These tasks are carried out by AI actors who are responsible for operating the AI system and working with others to regularly assess system output and impacts. AI actors in this category include system operators, domain experts, AI designers, users who interpret or incorporate the output of AI systems, product developers, evaluators and auditors, compliance experts, organizational management, and members of the research community. Test, Evaluation, Verification, and Validation (TEVV) tasks are performed throughout the AI lifecycle. They are carried out by AI actors who examine the AI system or its components, or detect and remediate problems. Ideally, AI actors carrying out verification Page 35 53 MST Al 100-1 AI RMF 1.0 and validation tasks are distinct from those who perform test and evaluation actions. Tasks can be incorporated into a phase as early as design, where tests are planned in accordance with the design requirement. • TEVV tasks for design, planning, and data may center on internal and external vali- dation of assumptions for system design, data collection, and measurements relative to the intended context of deployment or application. • TEVV tasks for development (i.e., model building) include model validation and assessment. • TEVV tasks for deployment include system validation and integration in production, with testing, and recalibration for systems and process integration, user experience, and compliance with existing legal, regulatory, and ethical specifications. • TEVV tasks for operations involve ongoing monitoring for periodic updates, testing, and subject matter expert (SME) recalibration of models, the tracking of incidents or errors reported and their management, the detection of emergent properties and related impacts, and processes for redress and response. Human Factors tasks and activities are found throughout the dimensions of the AI life - cycle. They include human -centered design practices and methodologies, promoting the active involvement of end users and other interested parties and relevant AI actors, incor- porating context -specific norms and values in system design, evaluating and adapting end user experiences, and broad integration of humans and human dynamics in all phases of the AI lifecycle. Human factors professionals provide multidisciplinary skills and perspectives to understand context of use, inform interdisciplinary and demographic diversity, engage in consultative processes, design and evaluate user experience, perform human -centered evaluation and testing, and inform impact assessments. Domain Expert tasks involve input from multidisciplinary practitioners or scholars who provide knowledge or expertise in — and about — an industry sector, economic sector, con- text, or application area where an AI system is being used. AI actors who are domain experts can provide essential guidance for AI system design and development, and inter- pret outputs in support of work performed by TEVV and AI impact assessment teams. AI Impact Assessment tasks include assessing and evaluating requirements for AI system accountability, combating harmful bias, examining impacts of AI systems, product safety, liability, and security, among others. AI actors such as impact assessors and evaluators provide technical, human factor, socio-cultural, and legal expertise. Procurement tasks are conducted by AI actors with financial, legal, or policy management authority for acquisition of AI models, products, or services from a third -party developer, vendor, or contractor. Governance and Oversight tasks are assumed by AI actors with management, fiduciary, and legal authority and responsibility for the organization in which an AI system is de - Page 36 54 NIST Al 100-1 AI RMF 1.0 signed, developed, and/or deployed. Key AI actors responsible for AI governance include organizational management, senior leadership, and the Board of Directors. These actors are parties that are concerned with the impact and sustainability of the organization as a whole. Additional AI Actors Third -party entities include providers, developers, vendors, and evaluators of data, al- gorithms, models, and/or systems and related services for another organization or the or- ganization's customers or clients. Third -party entities are responsible for AI design and development tasks, in whole or in part. By definition, they are external to the design, devel- opment, or deployment team of the organization that acquires its technologies or services. The technologies acquired from third -party entities may be complex or opaque, and risk tolerances may not align with the deploying or operating organization. End users of an AI system are the individuals or groups that use the system for specific purposes. These individuals or groups interact with an AI system in a specific context. End users can range in competency from AI experts to first-time technology end users. Affected individuals/communities encompass all individuals, groups, communities, or organizations directly or indirectly affected by AI systems or decisions based on the output of AI systems. These individuals do not necessarily interact with the deployed system or application. Other AI actors may provide formal or quasi -formal norms or guidance for specifying and managing AI risks. They can include trade associations, standards developing or- ganizations, advocacy groups, researchers, environmental groups, and civil society organizations. The general public is most likely to directly experience positive and negative impacts of AI technologies. They may provide the motivation for actions taken by the AI actors. This group can include individuals, communities, and consumers associated with the context in which an AI system is developed or deployed. Page 37 55 NIST AI 100-1 AI RMF 1.0 Appendix B: How AI Risks Differ from Traditional Software Risks As with traditional software, risks from AI -based technology can be bigger than an en- terprise, span organizations, and lead to societal impacts. AI systems also bring a set of risks that are not comprehensively addressed by current risk frameworks and approaches. Some AI system features that present risks also can be beneficial. For example, pre -trained models and transfer learning can advance research and increase accuracy and resilience when compared to other models and approaches. Identifying contextual factors in the MAP function will assist AI actors in determining the level of risk and potential management efforts. Compared to traditional software, AI -specific risks that are new or increased include the following: • The data used for building an AI system may not be a true or appropriate representa- tion of the context or intended use of the AI system, and the ground truth may either not exist or not be available. Additionally, harmful bias and other data quality issues can affect AI system trustworthiness, which could lead to negative impacts. • AI system dependency and reliance on data for training tasks, combined with in- creased volume and complexity typically associated with such data. • Intentional or unintentional changes during training may fundamentally alter AI sys- tem performance. • Datasets used to train AI systems may become detached from their original and in- tended context or may become stale or outdated relative to deployment context. • AI system scale and complexity (many systems contain billions or even trillions of decision points) housed within more traditional software applications. • Use of pre -trained models that can advance research and improve performance can also increase levels of statistical uncertainty and cause issues with bias management, scientific validity, and reproducibility. • Higher degree of difficulty in predicting failure modes for emergent properties of large-scale pre -trained models. • Privacy risk due to enhanced data aggregation capability for AI systems. • AI systems may require more frequent maintenance and triggers for conducting cor- rective maintenance due to data, model, or concept drift. • Increased opacity and concerns about reproducibility. • Underdeveloped software testing standards and inability to document AI -based prac- tices to the standard expected of traditionally engineered software for all but the simplest of cases. • Difficulty in performing regular AI -based software testing, or determining what to test, since AI systems are not subject to the same controls as traditional code devel- opment. Page 38 56 MST Al 100-1 AI RMF 1.0 • Computational costs for developing AI systems and their impact on the environment and planet. • Inability to predict or detect the side effects of AI -based systems beyond statistical measures. Privacy and cybersecurity risk management considerations and approaches are applicable in the design, development, deployment, evaluation, and use of AI systems. Privacy and cybersecurity risks are also considered as part of broader enterprise risk management con- siderations, which may incorporate AI risks. As part of the effort to address AI trustworthi- ness characteristics such as "Secure and Resilient" and "Privacy -Enhanced," organizations may consider leveraging available standards and guidance that provide broad guidance to organizations to reduce security and privacy risks, such as, but not limited to, the NIST Cy- bersecurity Framework, the NIST Privacy Framework, the NIST Risk Management Frame- work, and the Secure Software Development Framework. These frameworks have some features in common with the AI RMF. Like most risk management approaches, they are outcome -based rather than prescriptive and are often structured around a Core set of func- tions, categories, and subcategories. While there are significant differences between these frameworks based on the domain addressed — and because AI risk management calls for addressing many other types of risks — frameworks like those mentioned above may inform security and privacy considerations in the MAP, MEASURE, and MANAGE functions of the Al RMF. At the same time, guidance available before publication of this AI RMF does not compre- hensively address many AI system risks. For example, existing frameworks and guidance are unable to: • adequately manage the problem of harmful bias in AI systems; • confront the challenging risks related to generative AI; • comprehensively address security concerns related to evasion, model extraction, mem- bership inference, availability, or other machine learning attacks; • account for the complex attack surface of AI systems or other security abuses enabled by AI systems; and • consider risks associated with third -party AI technologies, transfer learning, and off - label use where AI systems may be trained for decision -making outside an organiza- tion's security controls or trained in one domain and then "fine-tuned" for another. Both AI and traditional software technologies and systems are subject to rapid innovation. Technology advances should be monitored and deployed to take advantage of those devel- opments and work towards a future of AI that is both trustworthy and responsible. Page 39 57 NIST AI 100-1 AI RMF 1.0 Appendix C: AI Risk Management and Human -AI Interaction Organizations that design, develop, or deploy AI systems for use in operational settings may enhance their AI risk management by understanding current limitations of human - AI interaction. The AI RMF provides opportunities to clearly define and differentiate the various human roles and responsibilities when using, interacting with, or managing AI systems. Many of the data -driven approaches that AI systems rely on attempt to convert or represent individual and social observational and decision -making practices into measurable quanti- ties. Representing complex human phenomena with mathematical models can come at the cost of removing necessary context. This loss of context may in turn make it difficult to understand individual and societal impacts that are key to AI risk management efforts. Issues that merit further consideration and research include: 1. Human roles and responsibilities in decision making and overseeing AI systems need to be clearly defined and differentiated. Human -AI configurations can span from fully autonomous to fully manual. AI systems can autonomously make deci- sions, defer decision making to a human expert, or be used by a human decision maker as an additional opinion. Some AI systems may not require human oversight, such as models used to improve video compression. Other systems may specifically require human oversight. 2. Decisions that go into the design, development, deployment, evaluation, and use of AI systems reflect systemic and human cognitive biases. AI actors bring their cognitive biases, both individual and group, into the process. Biases can stem from end -user decision -making tasks and be introduced across the AI lifecycle via human assumptions, expectations, and decisions during design and modeling tasks. These biases, which are not necessarily always harmful, may be exacerbated by AI system opacity and the resulting lack of transparency. Systemic biases at the organizational level can influence how teams are structured and who controls the decision -making processes throughout the AI lifecycle. These biases can also influence downstream decisions by end users, decision makers, and policy makers and may lead to negative impacts. 3. Human -AI interaction results vary. Under certain conditions — for example, in perceptual -based judgment tasks — the AI part of the human -AI interaction can am- plify human biases, leading to more biased decisions than the AI or human alone. When these variations are judiciously taken into account in organizing human -AI teams, however, they can result in complementarity and improved overall perfor- mance. Page 40 58 MST Al 100-1 Al RMF 1.0 4. Presenting AI system information to humans is complex. Humans perceive and derive meaning from AI system output and explanations in different ways, reflecting different individual preferences, traits, and skills. The GOVERN function provides organizations with the opportunity to clarify and define the roles and responsibilities for the humans in the Human -AI team configurations and those who are overseeing the AI system performance. The GOVERN function also creates mechanisms for organizations to make their decision -making processes more explicit, to help counter systemic biases. The MAP function suggests opportunities to define and document processes for operator and practitioner proficiency with AI system performance and trustworthiness concepts, and to define relevant technical standards and certifications. Implementing MAP function cat- egories and subcategories may help organizations improve their internal competency for analyzing context, identifying procedural and system limitations, exploring and examining impacts of AI -based systems in the real world, and evaluating decision -making processes throughout the AI lifecycle. The GOVERN and MAP functions describe the importance of interdisciplinarity and demo- graphically diverse teams and utilizing feedback from potentially impacted individuals and communities. AI actors called out in the AI RMF who perform human factors tasks and activities can assist technical teams by anchoring in design and development practices to user intentions and representatives of the broader AI community, and societal values. These actors further help to incorporate context -specific norms and values in system design and evaluate end user experiences — in conjunction with AI systems. AI risk management approaches for human -AI configurations will be augmented by on- going research and evaluation. For example, the degree to which humans are empowered and incentivized to challenge AI system output requires further studies. Data about the fre- quency and rationale with which humans overrule AI system output in deployed systems may be useful to collect and analyze. Page 41 59 NIST AI 100-1 Al RMF 1.0 Appendix D: Attributes of the AI RMF NIST described several key attributes of the AI RMF when work on the Framework first began. These attributes have remained intact and were used to guide the AI RMF's devel- opment. They are provided here as a reference. The AI RMF strives to: 1. Be risk -based, resource -efficient, pro -innovation, and voluntary. 2. Be consensus -driven and developed and regularly updated through an open, trans- parent process. All stakeholders should have the opportunity to contribute to the AI RMF's development. 3. Use clear and plain language that is understandable by a broad audience, including senior executives, government officials, non -governmental organization leadership, and those who are not AI professionals — while still of sufficient technical depth to be useful to practitioners. The AI RMF should allow for communication of AI risks across an organization, between organizations, with customers, and to the public at large. 4. Provide common language and understanding to manage AI risks. The AI RMF should offer taxonomy, terminology, definitions, metrics, and characterizations for AI risk. 5. Be easily usable and fit well with other aspects of risk management. Use of the Framework should be intuitive and readily adaptable as part of an organization's broader risk management strategy and processes. It should be consistent or aligned with other approaches to managing AI risks. 6. Be useful to a wide range of perspectives, sectors, and technology domains. The AI RMF should be universally applicable to any AI technology and to context -specific use cases. 7. Be outcome -focused and non -prescriptive. The Framework should provide a catalog of outcomes and approaches rather than prescribe one -size -fits -all requirements. 8. Take advantage of and foster greater awareness of existing standards, guidelines, best practices, methodologies, and tools for managing AI risks — as well as illustrate the need for additional, improved resources. 9. Be law- and regulation -agnostic. The Framework should support organizations' abilities to operate under applicable domestic and international legal or regulatory regimes. 10. Be a living document. The AI RMF should be readily updated as technology, under- standing, and approaches to AI trustworthiness and uses of AI change and as stake- holders learn from implementing AI risk management generally and this framework in particular. Page 42 60 This publication is available free of charge from: https://doi.org/1O.6028/NIST.AI. 100-1 NATIONAL INSTITUTE OF N[U.S. STANDARDS AND TECHNOLOGY DEPARTMENT OF COMMERCE CITY OF PORT ORCHARD Al Review Guidelines Introduction In addition to the City of Port Orchard's privacy, cyber, infrastructure, and data review protocols, the IT Department conducts an Al Review to ensure that the proposed Al system complies with the City of Port Orchard's Policy for Al systems, relevant privacy policies, and general technology acquisition review and processes. While not all steps of the City of Port Orchard's Al Review protocols may be necessary or appropriate for every project, the protocol provides the general review framework for any project. It is important to note that throughout the product lifecycle of all Al systems, there should be a consistent and continuous review of the technology through a human centric lens. Consistent review allows the City of Port Orchard to ensure that the Al system continues to provide value and protects against potential harm going undetected. What does an Al Review require? Not all Al systems, whether out of the box or customized, require a full Al Review.' Simple rule based systems, which rely on a series of hard coded conditional rules to produce one of multiple predefined outputs, may not be subject to the Al Review process. In contrast, algorithm based systems, which rely on complex logic to make predictions based on patterns in a set of training data, are more frequently subject to the Al Review process. Al systems should be reviewed on a regular basis (e.g., annually) Whether an Al system is subject to an Al Review also depends on the potential risk of the Al system in question. To classify the risk of an Al system, the IT Department conducts a risk threshold assessment before initiating an Al review of a given proposal (see the subsection "Step 2: Risk Analysis" within the section "Al Review Process: Step by Step" for more information on the risk threshold assessment). 1 It is important to note that vendors contracted under professional, personal, or general consulting agreements may be using AI based tools on the City of Port Orchard's behalf. When evaluating professional services agreements, determine if the vendor may be using AI systems to generate reports, analyze data, or provide insights and request information about such systems. Depending on the circumstances, the City of Port Orchard may wish to also review such third party AI systems in accordance with the same rules and policies described below. 62 See the chart below for examples of Al systems that do and do not require an Al Review. These examples are not exhaustive but rather are meant to guide and reinforce concepts. Al Review is required Al Review is optional 1. Predictive policing system 1. General website that impacts the deployment recommendations (e.g., of agency resources recommended videos on the 2. Identity recognition based on City of Port Orchard's one to many matching (e.g., YouTube channel) license plate reader, facial 2. Personalized notification recognition) system for City of Port 3. Al system that meets Orchard events organizational thresholds that 3. Strictly rule based logic (e.g., require an REP according to accounting software to procurement or contracts calculate taxes owed, a team motion based alarm system) 4. Automated decision making system that automates a decision which traditionally requires human review 5. Al systems that impact or integrate with infrastructure systems consistent with City of Port Orchard services (e.g., translation service for 311) Figure 1: An example list of Al systems that do and do not require an Al Review. 63 Al Review Framework: At a Glance The Al Review Framework guides the City of Port Orchard in reviewing Al systems during the public procurement process. The Framework outlines the actions that Al governance practitioners should take from the early project proposal stage to final approval and ongoing monitoring of the Al system. To understand how the Al Review process is adapted for Request for Proposals (RFP), refer to the section "Request for Proposals (RFP) Al Review Protocol". Procurement Request A City Department submits a procurement proposal for an Al tool �o Risk Analysis The City IT Department conducts a risk threshold analysis of the Al system Medium -high risk Algorithmic Impact Assessment The City IT Department conducts an algorithmic impact assessment, system is adjusted, and usage protocols are created 00 Impact assessment Al Fact Sheet Form (completed by (completed by Al IT Department) Vender) Figure 2: Al Review Framework. AI Review Framework Minimal risk 0 0 Approved Rejected Approved . 0 Final Review The City IT Department and/or the Al Committee complete final review Pre Launch Preparation 1. Post on Algorithmic Register 2. Train Users 3. Complete Data Usage Protocol (if needed} 1. Procurement Request or Existing Application Update: Ongoing Monitoring The Al system is monitored periodically for compliance and effectiveness Departments seeking to procure an Al system first engage the IT Department, to discuss their technology proposal. Departments should provide existing material related to the project, including details on project purpose, data collected and specific uses and benefits of automating a process, recommendation, or decision. 64 2. Risk Analysis: The IT Department, conducts an "Al Risk Threshold Analysis" to assess the risk of the proposed Al system and to determine if the project necessitates a full assessment. Step 2 below provides further guidance and examples of risk on a gradient scale from low risk to high risk.2 3. Assessment: The IT Department facilitates the following steps to evaluate the potential risks and benefits of the proposed Al system. The following steps should be taken in accordance with the City of Port Orchard's risk tolerance: a. Algorithmic Impact Assessment (AIA) Form: The business owning department(s) completes the required AIA Form, or equivalent. b. Al Factsheet: The vendor completes the required Al Factsheet. The IT Department may work with the vendor to obtain more details about the Al system. Equivalent information should be provided if the Al system is developed internally. If request involves a Request for Proposals (RFP): The vendor may not be determined at this time. Require potential vendor(s) to complete the Al Factsheet. See "Request for Proposals (RFP) Al Review Protocol" for guidance on RFP questions. Public Engagement: The City of Port Orchard should prioritize reducing barriers for public participation, particularly for those directly impacted by the Al system; such as, historically marginalized or disadvantaged communities. If the proposed Al system is considered of significant public interest, the Communications Specialist, or equivalent, conducts in person outreach, targeting communities with limited access to online comments (either due to language or internet access issues). Community feedback is then incorporated into the Data Usage Protocol. 2 It is important for the City of Port Orchard to understand and document its risk tolerance for AI systems. Key considerations that may be helpful for assessing an institution's risk tolerance include: What experience does City of Port Orchard have with Al? What are City of Port Orchard's existing risk management frameworks and practices? Who are the key stakeholders involved in City of Port Orchard's Al strategy? Are there any specific regulations or policies that will influence City of Port Orchard's use of AI? 65 The City of Port Orchard should prioritize reducing barriers for public participation, particularly for those directly impacted by the Al system, especially historically marginalized or disadvantaged communities. Final Review: Projects are reviewed by the IT department and Project Team members. The IT department provides assessment, approval/denial, and/or recommendations. At the discretion of the IT department a review may rise to the Al Committee. 4. Prelaunch Preparation: a. System Ownership Roles: The acquiring department or group, as the system owner, will designate an internal administrator who is responsible for maintaining expertise on the system use, capabilities, updates, and best practices. The internal administrator will keep the userbase and owning group apprised of relevant updates as well as provide mentorship and guidance to the City as to the use and direction of the system. b. Data Usage Protocol: Medium risk and high risk Al systems may necessitate a custom tailored Data Usage Protocol to govern the collection, access, processing, and sharing of data around the Al system to ensure that the project complies with the City of Port Orchard's Digital Privacy Policy. c. Al Inventory (Port Orchard Al Policy -Appendix B): The approved project proposal is added to the inventory of the Al systems deployed by the City of Port Orchard. The AIA Form and Al Factsheet are stored with this record. The IT Department regularly updates the Al Inventory as new Al systems are adopted and archives old systems as they are phased out of use. d. Training: Users of the approved Al system must receive training to properly deploy, operate, and maintain the technology. Training is typically provided by the vendor or other third party but is the responsibility of the system owner. 5. Ongoing Monitoring: Departments may be required to report annual metrics if defined in the Data Usage Protocol. Reports usually require metrics on data usage and project effectiveness. Al Review Framework: Step by Step The previous section provided a high level summary of the Al Review Framework. In this section, each of the seven steps that comprise the Al Review Framework are explained in greater detail. Step 1: Proposal An Al Review is triggered when a department or group in the City of Port Orchard submits a procurement request or technology review request for a technology involving an Al system. An Al Review can also be triggered when a vendor has made updates to a product's functionality or released a new version of the Al system. An Al Review can be formally initiated through the helpdesk site. For a consultation on potential projects, to initiate an informal "Al Risk Threshold Analysis", or to ask any questions, contact the IT Department through the helpdesk system. Step 2: Risk Analysis The City of Port Orchard will conduct a risk analysis to determine if a review is required; review will be overseen by the City IT Department. The City evaluates potential service delivery impact, operational impact, criticality, and principle based risk analysis 3 The City of Port Orchard aligns its approach to risk with the National Institute for Standards and Technology (NIST) Artificial Intelligence Risk Management Framework (Al RMF). While the NIST Al RMF does not necessarily describe exactly how to evaluate Al risk, the Al RMF defines characteristics of trustworthy Al and provides important guidance for organizations that seek to build governance and risk management processes for Al systems.4 While there are many methods to evaluating Al risk, the approach outlined below is designed to be easily implemented by public sector practitioners. The Al Risk Threshold Analysis is intended to be a practical tool that balances real world constraints with the need to adequately ascertain the risk of an Al system. In the Al Risk Threshold Analysis model, the impacted individual's inability to opt out of the use of the Al system and the severity of potential harm of the Al system are the two main factors for evaluating Al risk (see Figure 3).5 3https://www.bertelsmann-stiftung.de/fileadmin/files/BSt/Publikationen/GrauePublikationen/WKIO 2020 final.pdf " NIST AI Risk Management Framework: https://www.nist.gov/itl/ai-risk-management-framework and example applications: https://airc.nist.gov/Usecases. 5 The AI Risk Threshold Analysis model is partly inspired by the risk matrix found in AI Ethics Impact Group's "From Principles to Practice: An interdisciplinary framework to operationalise AI ethics". https://www.bertelsmann- stiftung.de/fileadmin/files/BSt/Publikationen/GrauePublikationen/WKI0 2020 final.pdf 67 Al Risk Threshold Analysis Pretrial algorithmic risk assessment . Search algorithm on City's website o Personalized informational • emails about City events Severity of potential harm Consider: How severe is the potential harm of the Al system if it is inaccurate, includes harmful bias, is not privacy -respecting, etc.? Components of potential harm Type of output: Decision, recommendation, forecast Level of oversight: No oversight, periodic oversight, individual oversight, individual action required Area of impact: Civil rights, health, employment, recreation. etc. Duration of impact: Long-term, mid-term, short-term Reversibility of impact: Irreversible, reversible Figure 3: The risk of an Al system can be estimated by identifying the impacted individual's inability to optout of the use of the Al system and the severity of potential harm posed by the system. The components of potential harm consist of the type of output, the level of oversight, the area of impact, the duration of impact, and the reversibility of the impact. In Figure 3, the risk level of each component of potential harm decreases from left to right. For example, for type of output, "decisions" are higher risk than "forecasts"; for level of oversight, "no oversight" is higher risk than "individual action required"; for area of impact, "impacts on civil rights" are higher risk than "recreational impacts". Although the Al Risk Threshold Analysis provides a systematic method for evaluating Al risk, estimating the risk of an Al system will vary for each context. For example, the distinction between a short term and middle term impact can be unclear and context dependent; should a duration of one month be considered a short term or a middle term impact? The answer will depend on the unique context in which the Al system is being deployed and on the values of the community (see "Step 4: Public Engagement"). Is an Al powered chatbot that leverages ChatGPT to provide residents with information about applying to social welfare programs a low risk or a high risk system? Again, the answer is highly context dependent and requires the contextual and subject matter expertise of the IT Staff, department specific authorized individual. It is possible for an Al system to simultaneously possess characteristics typically associated with both mid risk Al systems and characteristics typically associated with high risk Al systems. Categorizing these complex sociotechnical manifestations of risk is not always straightforward and requires the discretion of the practitioner conducting the risk analysis. When the risk level of an Al system is unclear, a good rule of thumb is to default to the higher risk categorization and conduct a full Al Review. 68 While estimating Al risk is not clearcut and requires the discretion of the IT Staff, department specific authorized individual, it can nevertheless be helpful when triaging risk to reference characteristics typically associated with low, mid, or high risk Al systems. The tables below feature characteristics that tend to be associated with low, mid, and high risk Al systems, along with examples of each risk archetype commonly found in the public sector. The tables below are not exhaustive and are intended to aid practitioners in thinking about Al risk. Low risk Al Systems Description Low risk: Al system that involves an opt in approach to a person being subject to the system. The system generates predictions but does not automate decision making and involves anonymized information used to provide general improvements to the City of Port Orchard. Notice is provided upon collection if any personal information is involved. The effects or impacts of the system are reversible, and typically short term. Characteristics • Data Collection: Notice is provided upon collection if any personal information is involved, as well as documentation to support the safe storage and handling of data. • Inferential: Al System provides analysis, insights, or predictions but these are for informational purposes only and are not tied to automated decision making. • Negligible impact on humans: The Al system is used for noncritical tasks with no negative material impact for humans. • Mundane applications: The Al assists with routine tasks like text completion. • Accuracy and Validity: Accuracy and validity metrics for the Al system are known. • Transparency: Access to appropriate levels of information based on the stage of the Al system's lifecycle is provided and tailored to the role or knowledge of individuals interacting with the Al system. • Explainable and interpretable: The meaning of the Al system's output(s) is understood in the context of its designed functional purpose. Examples of low risk Al • Software that generates a comprehensive profile of systems a client by aggregating data input. • A process that matches users to a basic administrative outcome such as time slots for appointments or next available client services specialist. • A system or tool that permits the operations of basic computer processes such as opening programs, sending electronic communications, autocorrecting, or using a calculator. • The use of generative Al for general research purposes, where the outputs are not included in public documents, policies, or decision making frameworks. Medium risk Al Systems Description Medium risk: Al system that involves identifiable information to provide targeted government services desired by the data owner with periodic oversight. Notice is provided at time of collection and often requires written consent. The system may have a short to medium term impact on quality of life factors. Characteristics • Data Collection: Notice is provided upon collection of sensitive personal information, but it is unknown how data is stored and handled. • Opt in/out consent: Users are automatically enrolled unless they actively choose to opt out. • Accuracy and Validity: Accuracy and validity metrics for the Al system are mostly known. • Transparency: Access to appropriate levels of information based on the stage of the Al Systems lifecycle is somewhat provided and tailored to general roles of individuals interacting with the Al system. • Explainable and interpretable: To a limited extent, the output(s) can be related back to inputs and model assumptions. 70 • Human Oversight: The Al system offers suggestions, insights or predictions, but humans retain final decision making power. • Economic impact: Has minor impact on workforce and economic opportunity. • Periodic oversight: Human monitoring occurs at intervals, not continuously. • Moderate impact: The Al system may temporarily impact quality of life, but has minimal long term risks, such as economic, legal, or reputational consequences. Examples of mid risk Al • Recruiting software that recommends relevant job systems openings to candidates based on their skills and experience, and final hiring decisions remain with human recruiters. • An Al model that evaluates loan applications based on financial data, but human loan officers make the final approval or denial decisions. • Marketing and advertising software that tailors ads to users' interests based on online behavior, while users retain control over their data and opt out options. • A tutoring software that tailors learning based on student data, however the student or the student's parents can choose the learning pathway and material difficulty. High risk Al Systems Description High risk: Al system that is potentially rights impacting or safety impacting within areas such as: critical infrastructure, biometrics, legal representation, and highly sensitive personal information traditionally kept hidden, like Social Security Numbers, credit card numbers, etc. The high risk system may automate decision making, have significant material impact on quality of life, and be subject to minimal or no human oversight. Characteristics • Compulsory: Users have no opt out option and are automatically subjected to the Al system. 71 • Data Collection: No notice or documentation is provided regarding collection, storage, and handling of personal data. • Automated Decisions: The Al directly drives decisions with minimal or no human input. • Minimal Oversight: Human monitoring is either absent or very infrequent. • Significant Impact: Decisions made by the Al system can profoundly affect quality of life, including: o Employment: Job opportunities, hiring or firing decisions, salary recommendations. o Healthcare: Diagnosis, treatment plans, eligibility for care. o Criminal Justice: Risk assessment, bail determinations, sentencing recommendations. o Public Safety: Conflict resolution, suspect selection, ticketing and fines, resource allocation. o Economic: Automated social services benefit distribution, audits. o Sensitive Data: The Al system likely processes highly sensitive personal data with significant consequences for misuse. o Financial: Automated spending which has the potential to violate liquidity and use of funds requirements, automated loan acceptance/denial. o Infrastructure: The reduction of internet bandwidth or power being inconsistent with the needs of the public. Examples of high risk Al • Generative Al that creates realistic, yet fabricated, systems videos or audio recordings, posing risks for misinformation and reputational harm. • A system that estimates individual risk factors for insurance, credit, employment, or healthcare without human oversight in decision making, potentially leading to bias and discrimination. • An autonomous weapons system that can choose and engage targets without human intervention. 72 • A biometric and facial recognition software that identifies individuals in real time based on facial features, raising privacy and potential bias concerns. • A system that recommends criminal sentences based on offender data and prior sentencing outcomes from historical case data. Step 3: Assessment During the assessment stage, the business owning department(s) fills out the required Algorithmic Impact Assessment Form for the IT Department, and the vendor completes the Al Factsheet. Higher risk projects require a Data Usage Protocol to guide how the project complies with the City of Port Orchard's Digital Privacy Policy. Lower risk projects may not need to undergo a full review. Algorithmic Impact Assessment Form The Algorithmic Impact Assessment (AIA) Form is completed by staff from the department procuring the system. The AIA Form consists of questions that are intended to capture information including, but not limited to: • Project objective: o Please clearly describe the project use case, the current process, and the desired outcome. o Which City of Port Orchard Department owns this system? o Who in the Department is responsible for this system? o Why does your department choose automation as an approach to this problem? What other approaches to solving this problem were considered (if any) and what led to choosing automation? • Vendor details: o Will the Al system be designed, developed, deployed, or maintained by vendors or third parties? o How can the agency test the Al system before it is put into use? • Transparency: o How do individuals receive notice in advance of interacting with the Al system? (For example, if a user is interacting with a chatbot, the system lets the user know they are talking to a chatbot instead of a human.) o How can third party auditors easily view the Al system's data to perform evaluations? o How could Al system operators or residents know if the system outputs an error? What ability will they have to correct or appeal an error? • Equity: o What individuals and communities will interact with the Al system? For example, is the algorithm used on the general population (technology used 73 in many public areas) or a specific group (e.g., children in a school program, a single neighborhood)? o How likely is it that the Al system impacts children under the age of 18? o Does this use, and the information/decisions generated by the Al system, impact an individual's rights or freedoms (e.g., if the Al system helps determine if a suspect can be put on bail or must remain in jail)? o Does this use, and the information/decisions generated by the Al system, impact an individual's economic status (e.g., if the Al system helps determine if an individual can apply to affordable housing)? o Does this use, and the information/decisions generated by the Al system, impact an individual's health, healthcare, wellbeing (e.g., if the Al system helps determine an individual's likeliness for colon cancer)? o Do decisions from the Al system impact the environment? (e.g., potential impact to carbon emissions, high tech waste)? o What issues could arise if the Al system is inaccurate? • Human oversight o Please describe the level of autonomy of the Al system. • System operates automatically with no human intervention • System operates automatically with occasional retrospective reviews by humans ■ System operates automatically with opportunity for humans to override any individual action ■ System produces recommendations but cannot act without human intervention o If there is human intervention in the Al system, is it by the vendor, agency department/office, or both? o Please list the agency roles/divisions that will be "touching" the system or managing the deployment and use of the Al system. o How does the department provide training and resources for personnel to help them develop the skills they need to effectively operate the Al system? o In the event the Al system does not work or is deemed to be inaccurate, what back up measures are in place to continue providing services? In other words, can the City continue to provide the service without the Al system, and how would it do that? • Accessibility o Have you considered how the Al system integrates and interacts with commonly used assistive technologies (e.g., screen readers, voice recognition software, etc.)? o Have you considered how users of diverse abilities will interact with the user interface? o How will feedback be collected by individuals with disabilities regarding the system? 74 o How will feedback from individuals with disabilities be implemented into the system? o Where will accessibility features and resources be documented and readily available? o Will there be specialized training for individuals with accessibility needs? • Liability o Who in City of Port Orchard will ultimately be accountable and responsible if the system fails to operate as intended? o Who in City of Port Orchard has the authority to stop or limit the Al system's use? o If a vendor fails to meet contractual obligations, what are the alternative options that exist to ensure there is no loss of service? (IT Incident Response Plan) Port Orchard Al Vendor Factsheet The Al Factsheet is completed by the vendor of the Al system and captures basic facts about the Al system. The Al Factsheet enables the IT Department, and potentially the Al Committee, to better understand the technical details of the Al system and ultimately assess the risks and benefits it presents. The Al Factsheet is intended to capture information including, but not limited to: • Training data • Testing data • Input and outputs • Performance metrics • Optimal conditions • Poor conditions • Bias The IT Department will work with the vendor as needed to obtain necessary technica details about the Al system. Step 4: Public Engagement If the proposed Al system presents a significant potential risk or is of significant public interest (for example, the use of Flock Cameras for Police), the Communications Specialist conducts in person outreach, targeting communities with limited access to online comments (either due to language or internet access issues). If the proposed Al system presents a significant potential risk or is of significant public interest (for example, the use of Flock Cameras for Police), the Communications Specialist, or equivalent, conducts in person outreach, targeting communities with limited access to online comments (either due to language or internet access issues). Community feedback is then incorporated into the Data Usage Protocol. 75 The City of Port Orchard should prioritize reducing barriers for public participation, particularly for those directly impacted by the Al system; such as, historically marginalized or disadvantaged communities. Public engagement can occur online, in person, and in the built environment. Below are examples of formats for public engagement: • Online o Interactive website portal: Create a dedicated website where citizens can submit feedback on specific Al initiatives, participate in surveys, and engage in discussions. o Social media town halls: Host live Q&A sessions on social media platforms like Facebook, Instagram, Reddit, or Bluesky where experts discuss Al and answer public questions. o Online forums and communities: Create dedicated online forums or communities, possibly on sites like Linkedln or Reddit, focused on Al policy and development. • In person o Public workshops and town halls: Organize in person events where citizens can learn about Al, hear from experts, and discuss their concerns and suggestions. In addition, create interactive activities like brainstorming sessions to gain feedback. o Community outreach programs: Partner with local organizations, libraries, and community centers to host Al education and feedback sessions. o Citizen advisory boards: Establish a diverse advisory board of citizens to provide ongoing feedback on Al development and policy. o Focus groups and interviews: Conduct targeted focus groups and interviews with specific demographics or stakeholders to gather in depth feedback on specific Al applications or concerns. • Built environment o Interactive kiosks and installations: Install interactive kiosks in public spaces like libraries, parks, or government buildings where citizens can learn about Al and provide feedback through surveys, polls, or open ended questions. o "Living labs" for testing Al applications: Designate specific areas or neighborhoods as "living labs" where citizens can experience and provide feedback on prototype Al applications in real world settings. • Additional considerations o Accessibility and inclusivity: Ensure all feedback channels are accessible to people with disabilities and diverse backgrounds. Use multiple languages, alternative formats, and assistive technologies. 76 o Transparency and communication: Clearly communicate the purpose of collecting feedback, how it will be used, and how citizens can stay informed about the process. o Data privacy and security: Implement data security measures to protect citizen privacy and ensure feedback is handled ethically and responsibly. Step 5: Review After the necessary documentation and public engagement has been completed, the IT Department reviews the proposal and provides final approval or rejection of the proposal. Depending on the level of risk presented by the proposal, this step may involve gaining approval from the Al Committee, the Mayor, City Council, or equivalent. The review may require input from the City Attorney's Office or other relevant departments on a case by case basis. Based on the review by IT Department or Al Committee, in some instances, the proposal may need to be revised before being approved. Step 6: Pre Launch Preparation Following approval, relevant documentation will be published in the technology documentation repositories including the Application Catalog (Maintained by the IT Department), and the Al Technology Registry (Port Orchard Al Policy - Appendix B). Prior to implementation, relevant parties (e.g., staff in business owning department) are given training to properly deploy, operate, and maintain the technology. Training is often provided by the vendor or other third party but is the responsibility of the system owner. Step 7: Ongoing Monitoring The Data Usage Protocol for a given technology proposal may require that the business owning department of the Al system submits an Annual Usage Report. The report is typically 1-2 pages, drafted by the applicable department(s) and details: 1. Project summary 2. Required performance metrics as defined in the Data Usage Protocol (e.g., accuracy, effectiveness, cost) 3. Future plans for the technology initiative (e.g., project expansion, shift in usage) 77 Protocol for a Request for Proposals Prior to issuing a Request for Proposals (RFP) or similar procurement process, departments should review the proposed solution to discern if it includes an Al component. If so, the person requesting the RFP should file an Al Acquisition request following these guidelines, or reach out to your department's Al Committee liaison to trigger an Al Review. RFPs should incorporate questions that demand transparency around how the Al model works, clarify what protocols for human oversight are in place, and confirm that there are mechanisms for user review. In consideration of the City of Port Orchard's public records obligations and transparency commitments, all vendors subject to the use of the protocols outlined in this document through the RFP process should be preemptively informed of City of Port Orchard's intended or required disclosure practices around bid documentation. Use of the RFP Protocol This document should be used in conjunction with the Al Factsheet, Al Review, and Vendor Agreement during the public procurement process. This document assumes the following RFP process: 1) Establish minimum requirements Al vendors must meet to be considered for Agency/Project bid and local procurement laws. 2) Publish Al Factsheet as part of the Agency's RFP solicitation documentation. Require vendors to provide a completed Al Factsheet as part of their formal bid or proposal. 3) Subject Matter Expert (SME) will ask vendor additional questions provided in this document. An illustrative example of assessment questions and a scoring model of how to evaluate answers to these questions for an Al system is below. Recommended point values on a scale of zero to five are also provided for each category. Each section of assessment questions includes a "nontechnical" question that asks the vendor to provide an answer in plain language that can be easily understood by a nontechnical audience. Any scoring methodology associated with evaluation of RFP bids is determined independently by City of Port Orchard in alignment of City of Port Orchard's stated principles and priorities. Please use this as a guide or rubric for best practices. It is the vendor's responsibility to be responsive to these questions, and it is the SME's responsibility to ensure that the vendor provides meaningful answers to the assessment questions. 78 Assessment Questions 1. System Overview • Brief summary of the Al system. • Purpose of the Al system, the intended use case, and users. • Relevant context to the technology and maturity of the vendor. • What is the City of Port Orchard policy on data collection, storage, and distribution? • Training materials and implementation plan. • Nontechnical: Can you provide a nontechnical overview of how your Al system operates and its key functionalities? 2. Data Training and Model Description • How was the Al system trained, and what data was used? • How often is data added to the training set? • What data was used to test system performance? • What conditions has the system been tested under? • Provide a general description of the model(s) used. • Nontechnical: In layperson's terms, can you explain how your Al system learns information and what kind of data it has been trained on? 3. System Operations • How often are the models updated for users? • Will the user have a choice in moving to the updated model or staying on the current model? • Where is prompt and output data stored? Is this information used for future model versions? • Do operators require specific education or certification to use the system? • Nontechnical: From a user's perspective, can you clarify where data is stored, as well as the process is for receiving updates or choosing to stay with a current model version? 4. Performance Evaluation • How was the accuracy and effectiveness of the system measured? • What metrics were used, and why? • What is the range of accuracy of the Al system, and how does it vary depending on the data? • What is the system optimizing for and under what constraints? • Nontechnical: In simple terms, how well does your Al system perform, and what aspects do its performance metrics prioritize? 5. Ethical Considerations 79 • What biases does the tool exhibit, and how does it handle that bias? • Does the vendor report bias or justify why no bias would be present? • How does the tool prevent or reduce harm to the end user? • Nontechnical: How do you ensure that your Al system treats all individuals and groups fairly, without any unintended biases? 6. System Reliability • How does the Al system handle outliers? • Do overwritten decisions feed back into the system to help calibrate it in the future? • What conditions does the model perform best under? • What conditions does the model perform poorly under? • What are the limitations of the Al system? • What expertise does this Al system require for operation, debugging, modification, and troubleshooting. • Nontechnical: Can you explain, in nontechnical terms, how your Al system deals with unusual cases or incorrect predictions? 7. Interpretability and Explanation • How does the Al system explain its predictions? • Are the outcomes of the Al system understandable by subject matter experts, users, impacted individuals, and others? • Nontechnical: Can you share examples or scenarios illustrating how the Al system communicates its predictions in a way that is easy to understand for laypeople? 8. Monitoring and Correction • How is the Al tool monitored to identify any problems in usage? • Can outputs (recommendations, predictions, etc.) be overwritten by a human? • Do overwritten outputs help calibrate the system in the future? • Nontechnical: For end users, how can they be involved in monitoring and correcting any issues with the Al system? 9. Studies and Transparency • Have the vendors or an independent party conducted a study on the bias, accuracy, or disparate impact of the system? • If yes, can the City of Port Orchard review the study? • Include methodology and results. • Is the data used to train the system that is representative of the communities it covers? 80 • Nontechnical: Can you provide examples of studies conducted to ensure fairness and accuracy, and how transparent are these studies? 10. User Interaction and Feedback • How can the City of Port Orchard and its partners flag issues related to: bias, discrimination, or poor performance of the Al system? • How is the Al tool made accessible to people with disabilities? • Has it been assessed against any usability standards, and if so, what was the result? • What other human factors, if any, were considered for usability and accessibility of the system? • Nontechnical: How can users easily provide feedback on any issues they encounter with the Al system, and what measures have been taken to ensure accessibility for all users? Scoring Model 1. System Overview • 5 (Excellent): A comprehensive, nontechnical overview that effectively communicates the Al system's purpose and functionality. • 4 (Good): A clear and concise summary, providing a basic understanding of the Al system. • 3 (Average): A brief overview but lacks clarity in conveying the system's purpose. • 2 (Below Average): Limited information that does not effectively convey the system's purpose. • 1 (Poor): No overview or insufficient information to understand the context. 2. Data Training and Model Description • 5 (Excellent): Detailed information on training data, model architecture, and transparency on model usage. • 4 (Good): Clear explanation of the training process and model description. • 3 (Average): Basic information on training data and models without much detail. • 2 (Below Average): Limited information on training data and model description. • 1 (Poor): No information or inadequate details regarding training data and model. 3. System Operations • 5 (Excellent): Regular model updates with clear communication and user friendly options. • 4 (Good): Frequent updates with communication on changes, providing user choice. • 3 (Average): Regular updates but lacking clear communication or user choice. 81 • 2 (Below Average): Infrequent updates with unclear communication and no user choice. • 1 (Poor): No updates or communication about the system's status. 4. Performance Evaluation • 5 (Excellent): Comprehensive metrics, clear justifications, and superior performance compared to other vendors. • 4 (Good): Effectively explained metrics with justified performance in line with industry standards. • 3 (Average): Basic metrics explanation with average performance. • 2 (Below Average): Limited metric explanation with subpar performance. • 1 (Poor): No metric explanation or poor performance without justification. 5. Ethical Considerations • 5 (Excellent): Thorough identification and handling of biases with transparent reporting. • 4 (Good): Clear recognition and handling of biases with transparency. • 3 (Average): Basic acknowledgment of biases without much transparency. • 2 (Below Average): Limited recognition or handling of biases. • 1 (Poor): No acknowledgment or handling of biases. 6. System Reliability • 5 (Excellent): Robust handling of outliers, effective calibration, and adaptability to corrections. • 4 (Good): Efficient handling of outliers and adaptability to corrections. • 3 (Average): Adequate handling of outliers with some adaptability to corrections. • 2 (Below Average): Limited handling of outliers and minimal adaptability to corrections. • 1 (Poor): No handling of outliers or adaptability to corrections. 7. Interpretability and Explanation • 5 (Excellent): Clear and understandable explanations that cater to both experts and general users. • 4 (Good): Comprehensible explanations for predictions. • 3 (Average): Basic explanations that may lack clarity. • 2 (Below Average): Limited explanations that are often unclear. • 1 (Poor): No explanations provided or entirely incomprehensible. 8. Monitoring and Correction • 5 (Excellent): Robust monitoring, efficient correction mechanisms, and clear user involvement. 82 • 4 (Good): Efficient monitoring and correction mechanisms with user involvement. • 3 (Average): Adequate monitoring with basic correction mechanisms and user involvement. • 2 (Below Average): Limited monitoring and correction mechanisms with minimal user involvement. • 1 (Poor): No monitoring or correction mechanisms, and no user involvement. 9. Studies and Transparency • 5 (Excellent): Independent studies, transparent methodologies, and representative training data. • 4 (Good): Third party studies with transparent methodologies. • 3 (Average): Some transparency in studies and methodologies. • 2 (Below Average): Limited transparency in studies and methodologies. • 1 (Poor): No studies or transparency in methodologies. 10. User Interaction and Feedback • 5 (Excellent): User friendly feedback mechanisms, high accessibility, and positive usability assessment. • 4 (Good): Effective feedback mechanisms with good accessibility. • 3 (Average): Adequate feedback mechanisms and accessibility features. • 2 (Below Average): Limited feedback mechanisms and basic accessibility. • 1 (Poor): No feedback mechanisms or accessibility features. Use case specific questions Practitioners may want to include additional questions based on the specific use case for the Al system. Below are a few examples of use case specific questions. The lists below should be used as a starting point for ideas for questions that would be helpful to ask a vendor. Practitioners should review the Use Case Template to build out the Al use case. The Use Case Template is a helpful tool for framing Al use cases in the public sector. Practitioners should resist the urge to tie a use case to a solution and instead be solution agnostic. Use Case: Translation Based Al Systems • Please provide a list of all languages your solution supports for live speech translation, beyond the mandatory languages specified in Attachment A. Additionally, describe how often you introduce new languages to your platform and outline the process for these additions. • Does your solution offer text to speech functionality or other methods to accommodate American Sign Language speakers in both in person and virtual meetings? If so, please describe the features and functionalities that enable this 83 accommodation, including any limitations or requirements for optimal performance. • How do you measure the performance of your solution in terms of translation, speech to text, and text to speech for English, Spanish, Tagalog, Mandarin, and Vietnamese? Please specify the metrics used (e.g., BLEU score, human ratings, etc.) and the conditions under which these measurements were taken. Based on these metrics, what is the performance of your solution for the specified languages? • Please provide/describe any third party evaluations or benchmarks that have been conducted on your solution's translation capabilities. Include details on the evaluation process, criteria, results, and any subsequent improvements made to the solution based on these evaluations. • What is your solution's average time delay between speech input and translation output? • How does the solution handle rapid speech or overlapping conversations? • How can the solution adapt to idiomatic expressions, cultural references, or local slang? Use Case: ADA Compliance and Accessibility Considerations • How well does your solution adapt to blurring, obstruction, poor lighting or any conditions that may lead to misclassification? • Does your product offer built in accessibility features for users with visual, auditory, motor, or cognitive disabilities? Please describe these features in detail. • Has your product been tested for compatibility with common assistive technologies like screen readers, voice recognition software, and alternative input devices? • Has your product undergone accessibility testing with users who have disabilities? If so, please share the testing methodology and key findings. How can users with disabilities provide feedback about the product's accessibility? Are there dedicated channels or support options for such concerns? 84 Policy Review and Approval History Approval Version Content Contributors Date v 1.0 Initial Reviewer(s): Sean Dunham -IT Manager TBD Draft v.1.1 2nd Draft Reviewer(s): Everyone 85 Data, Input, and Output Input Data Type(s) Output Data Type(s) Training Data Description Training Data Temporal Relevance Testing/Validation Data Description PII/Sensitive Data Handling Data Retention Policy Performance and Conditions Primary Performance Metric Reported Metric Score Latency/Response Time Optimal Operating Conditions Poor Operating Conditions/Failure Modes Confidence Threshold Risk and Bias Assessment Known Bias Categories Mitigation Strategies Explainability (XAI) Required Human Oversight Audit Log Capabilities Compliance/Regulatory Hardware and Support Required Hardware/OS (On -Premises) Integration API/SDK Used Support Channels Service Level Agreement (SLA) Uptime Vendor Response Vendor Response Vendor Response Vendor Response 87 Required Vendor Detail What kind of data does the system ingest? (e.g., Text Documents, Images/PDFs, Numeric Datasets, Sensor Readings) What is the system's primary output? (e.g., Text Summary, Numeric Score/Risk Rating, Classification Label, Automated Workflow Trigger) Describe the dataset(s) used to train the model (source, size, domain). What is the age range of the training data? (e.g., 2010-2023) Describe the dataset(s) used for validation/testing, ensuring it is distinct from training data. Does the system process or store PII (Personally Identifiable Information), Confidential, or Protected Data? (Yes/No) How long is City data stored on the vendor's servers? (If SaaS) Required Vendor Detail The key metric the model is optimized for (e.g., Accuracy, Precision, F1 Score). The measured score on the validation dataset (e.g., 95% Accuracy). Typical time from input submission to output generation. What conditions lead to the best performance? (e.g., Structured PDF format, high -resolution images, input text $\It 500$ words). What conditions cause performance degradation or failure? (e.g., Handwritten text, input documents in languages other than English, missing required fields). Does the system have a confidence score threshold? If so, what is the action taken when the score falls below that threshold? Required Vendor Detail Has the model been tested for bias based on demographic categories? If so, which categories show a difference in performance? (e.g., Geographic bias, Socioeconomic bias) What steps have been taken to reduce or eliminate identified biases? Is the model output explainable? (i.e., Can the system provide a clear, auditable reason for its decision/score?) Is human review required before the Al's output is made final or public? (Yes/No) Does the system log every input, output, and decision for audit purposes? What relevant industry or government standards does the system comply with? (e.g., SOC 2, ISO 27001, FedRAMP, HIPAA) Required Vendor Detail Specify minimum CPU, RAM, GPU (if applicable), and required OS versions. Name the specific API or SDK used for integrating with City systems. (Select One or More) Phone, Email, Ticketing System, 24/7 Monitoring Guaranteed percentage of system uptime per contract. 88 IAlaorithmic Impact Assessment (AIA) Form This form is to be completed by staff from the business owning department procuring the Al/algorithmic system. The goal of this assessment is to capture information necessary to evaluate the potential benefits, risks, and implications of deploying an algorithmic system within the City of Port Orchard. Please fill out each tab below with as much detail as ssible. 89 City of Port Orchard Data Classification Guidelines Effective Date: Data Classification Categories All City of Port Orchard data should be categorized into one of the following levels, based on its sensitivity and the potential impact of its unauthorized disclosure, modification, or destruction. • Public Data: Data that is not protected by law or regulation and is intended for public disclosure. The unauthorized release of this data would not cause any significant adverse impact to the city, its residents, or employees. o Examples: City meeting minutes and agendas, press releases, city budgets, public - facing website content, and general government reports. • Internal Use Only: Data that is not highly sensitive but is intended for use by City of Port Orchard employees or authorized contractors. This data is not protected from public disclosure by law, but its integrity and availability should be protected. o Examples: Inter -Agency memorandums, preliminary drafts, non-public employee contact lists, and internal project documents. • Confidential Data: Information that is protected from disclosure by federal or state law, or by a contractual agreement. Unauthorized disclosure could cause moderate to severe harm, including legal sanctions, financial loss, or a breach of privacy. o Examples: Most employee personnel records, sensitive financial data, and some law enforcement records. This category includes data specifically protected by laws like the Washington State Public Records Act (RCW 42.56) and other state and federal statutes. • Highly Confidential Data: The most sensitive data. Unauthorized disclosure, modification, or destruction of this data could result in severe harm, including threats to public health and safety, legal sanctions, catastrophic financial loss, or a violation of an individual's rights. o Examples: Personally Identifiable Information (PII), such as Social Security numbers, driver's license numbers, and financial account information combined with a person's name; Protected Health Information (PHI); and certain law enforcement data that could compromise ongoing investigations or endanger individuals. Roles and Responsibilities Establishing clear roles and responsibilities is crucial for a successful data classification program. • Data Owner: The department head or a designated employee who is responsible for the data. They are accountable for classifying their data, ensuring it's handled according to policy, and approving access requests. • Data Custodian: The IT department or another designated team responsible for implementing and maintaining the technical controls and security measures for the data, such as storage, backup, and access management. • Data User: Any City of Port Orchard employee or authorized third party who accesses and uses city data. All users are responsible for handling data in accordance with its classification level. Handling Procedures The following are general guidelines for handling data based on its classification. These apply to both physical and electronic data. Classification Handling Procedures No specific handling requirements beyond general security practices. May Public Data be posted on the city's website or shared with the public without restriction. Internal Use Only Store on approved city network drives. Do not post on public websites or share with unauthorized external parties. Must be stored in a secure location (locked file cabinet or encrypted Confidential Data network drive). Electronic data must be encrypted when transmitted. Access must be restricted to individuals with a legitimate business need. Requires the most stringent security measures. Must be stored in a Highly Confidential secured and encrypted location. Transmitted only via secure, encrypted Data methods. Access is strictly limited and must be audited regularly. Paper documents must be crosscut shredded when no longer needed. Policy Review and Approval History Version Content Contributors Approval Date Reviewer(s): Sean Dunham -IT Manager, v 1.0 Initial Draft August 21,2025 Jake Langston -IT Support Specialist 91 ORCHARD Establishing a Framework for Responsible Al Usage 92 Purpose of the Al Committee Why Al policy matters for Port Orchard Alignment with city goals and ethical standards 93 The Call to Action Initial Invitations: January 17, 2025 First Meeting: April 24, 2025 Initial Focus • Introduction to basic Al concepts and terminology • Review of educational videos to establish a shared baseline • Discussion of potential municipal use cases and immediate needs 94 Cross Departmental Collaboration • Finance: Noah Crocker (Finance Director), Rebecca Zick (Deputy Finance Director), Kori Pearson (Acct. Asst. III/IT Specialist) • Information Technology: Sean Dunham (IT Manager), Jake Langston (IT Specialist) • Administration: Debbie Lund (HR Director), Brandy Wallace (City Clerk), Jenine Floyd (Deputy City Clerk) • Operational Leadership: Denis Ryan (Public Works Director), Nick Bond (DCD Director), Jim Fisk (Principal Planner), Caden Cucciardi (Asset Management Technician) • Public Safety: Alan Iwashita (Deputy Police Chief) 95 Policy Development Timeline Subcommittee to Final Draft Policy Subcommittee formed July 2025 First Meeting August 2025 Urgency: The committee formed an Al policy subcommittee immediately upon deciding a formal policy was needed "asap" Drafting Process: The subcommittee worked through several iterations (v0.1 through v0.6) between July and December 2025 Finalization: Version 1.0 was completed on December 16, 2025 Distribution: The final draft was provided to HR for citywide distribution to employees and union representatives for comments 96 Establish clear Al governance framework Ensure transparency and accountability Promote responsible innovation Protect privacy and security 97 Guiding Values for Al Use Innovation: Commit to responsibly exploring Al to improve community services Transparency & Accountability: Ensure Al systems are compliant Core Principles with all laws and that documentation is available publicly Bias Reduction: Actively evaluate systems to address potential algorithmic or human bias Data Privacy: Apply standard operating procedures to reduce privacy risks throughout the Al lifecycle 98 Ensuring Accuracy and Accountability • Human Oversight (HITL): All Al outputs must be reviewed by a human prior to use in any official City capacity • Attribution Requirement: Substantive use of Al generated content in final products requires clear attribution • Example Attribution: "Some materia/in this brochure was generated using [Al System] and was re vie wed for accuracy by [Department/Group]" • Responsibility: Reviewers must ensure output is accurate and free of discrimination 99 Data Security and Privacy Protecting City Information Restricted Data: Users shall not submit "Confidential," "Highly Confidential," or non-public information to Al systems outside City control Model Training: City data, including prompts, may not be used to train or tune Al models outside City control Prohibited Tools: Technologies that cannot prevent City data from contributing to their language models are barred from use 100 Formal Approval Process • Approved Registry: Microsoft Copilot is currently the primary approved Al tool New Requests: Department directors must request new tools through the IT request process; the Al Committee may provide final approval/denial Automatic Suspension: If existing software adds Al capabilities, they may be suspended by IT pending their formal review, and potentially by the Al Committee Consequences: Noncompliance may result in disciplinary action, up to and including termination 101 Collaborative Refinement • Distribution Results: Comments were received from various departments following the HR distribution • Resolution: All technical and procedural concerns were addressed by the IT department • Current Status: The policy is now ready for full implementation, ensuring legal and operational safeguards are in place 102 6 Procurement Request Risk Analysis A City Department submits a The City IT Department conducts a procurement proposal for an Al risk threshold analysis of the Al tool system Medium► -high risk Minimal risk 0 Approved Algorithmic Impact Assessment The City IT Department conducts an algorithmic impact Rejected assessment, system is adjusted, and usage protocols are ■ created Ir ® Q Q Approved • Impact assessment Al Fact Sheet Final Review Pre Launch Preparation Ongoing Monitoring Form (completed by (completed by Al The City IT Department and/or the 1- Post on Algorithmic Register The Al system is monitored IT Department) Vendor) Al Committee complete final 2- Train Users periodically far compliance and review 3. Complete Data Usage Protocol effectiveness (if needed) 103 Implementation and Support • Final policy adoption • Distribution of an Al Users Guide and training materials • Use Case discussions and assessments • Establishing departmental standards for "Human -in -the -Loop" (HITL) reviewers 104