OMB M-25-21: Federal agency use of AI
Tells federal agencies how to adopt and govern AI: each agency names a Chief AI Officer, large agencies publish AI strategies, and 'high-impact' AI uses (those affecting rights or safety) must meet minimum risk practices such as testing and human oversight, or be stopped. It replaced the Biden-era M-24-10.
At a glance
In force
- Jurisdiction
- United States
- Type
- Guidance
- Official title
- OMB Memorandum M-25-21: Accelerating Federal Use of AI through Innovation, Governance, and Public Trust
What this means for you
If you build AI
- If you sell to federal agencies, expect requests that support their minimum practices for high-impact AI, such as testing, impact assessments and monitoring.
If you deploy AI
- Applies to federal agencies, not private companies. Agencies publish AI use case inventories and compliance plans you can read.
If you use AI
- When a federal agency uses AI in decisions that significantly affect you, this memo requires minimum safeguards, including human oversight and ways to seek remedies.
Key dates
No key dates recorded.
Status history
- Apr 3, 2025In forceIssued by OMB; replaces M-24-10. Still listed (not rescinded) on OMB's memoranda page as of 2026-10-04Source
In the Drops
No Drop has covered this law yet.