21-06 | Agencies regulating private sector AI: apply the ten principles and keep the AI regulatory plan current | 0 |
21-2a | Publish an agency AI Strategy within 180 days (CFO Act agencies) | 0 |
21-2b | Share AI code, models and data across government and release code as open source where practicable | 0 |
21-2d | Procure AI so that the agency keeps its data rights, can measure performance and sustains competition | 0 |
21-2e | Build an AI-ready workforce: upskill staff, recruit experienced AI talent, track talent needs | 0 |
21-3a1 | Designate a Chief AI Officer within 60 days with the stated authority and duties | 0 |
21-3a2 | Convene an agency AI Governance Board within 90 days (CFO Act agencies) | 0 |
21-3b | Governance responsibilities: empower AI leaders, submit compliance plans, update IT, data, cybersecurity and privacy policies, adopt a generative AI policy, inventory use cases ann | 0 |
21-3c | Chief AI Officer Council and OMB machinery | 0 |
21-4a | High-impact AI: determine and document high-impact use, implement the minimum practices within 365 days or discontinue, pilots, waivers and public reporting | 0 |
21-4b1 | High-impact AI: pre-deployment testing with risk mitigation plans | 0 |
21-4b2 | High-impact AI: a documented AI impact assessment before deployment, updated through the lifecycle | 0 |
21-4b3 | High-impact AI: ongoing monitoring, operator training, human oversight and fail-safes, remedies and appeals, and feedback from users and the public | 0 |
22-3b | Update acquisition procedures within 270 days: review planned AI acquisitions, convene cross-functional teams, use IP terms | 0 |
22-3cd | Maximise American-made AI and protect privacy in AI acquisition with early SAOP involvement | 0 |
22-3e | Protect IP rights and government data: scope licences, keep operating components available, limit vendor retention, ban training on agency data without consent | 0 |
22-3h | Decide when to require disclosure of AI used in contract performance | 0 |
22-4ab | Acquisition lifecycle: identify requirements with the cross-functional team, make an early high-impact determination, research the market and use performance-based techniques | 0 |
22-4c | Solicitation development: disclose high-impact use and documentation demands, build in anti-lock-in provisions and IP and data terms | 0 |
22-4d | Selection and award: test proposed solutions, re-evaluate risk, and include the required contract terms | 0 |
22-4e | Contract administration and closeout: monitor performance, manage new features and plan for transition | 0 |
26-04a | LLM procurement: contractual requirements on the Unbiased AI Principles in every new solicitation, existing contracts modified before option exercise | 0 |
26-04b | Update procurement policies by 11 March 2026 with a user reporting process for principle-violating outputs | 0 |
26-04c | LLM transparency: minimum vendor documentation, enhanced documentation where needed, and materiality for termination | 0 |