Federal AI Use Case Inventory for 2025 Is Imminent
The White House Office of Management and Budget is preparing the federal government's 2025 AI use case inventory for public release. Federal CIO Greg Barbaccia said, "OMB is compiling publicly posted AI use case inventory submissions and will release a consolidated Federal resource on GitHub soon."
This will be the latest annual inventory and the first issued during the current Trump administration, which has pressed agencies to expand AI adoption in pursuit of efficiency gains.
Deadlines, delays, and what changed
Agencies faced a Nov. 4 submission deadline and a Dec. 2 posting deadline for their 2025 inventories. Those dates overlapped with the government shutdown, which pushed timelines back. In December, Barbaccia indicated the slip wouldn't be a major delay.
Unlike prior years, the detailed guidance to agencies has not been posted publicly. Some agencies have already published 2025 inventories on their sites, with the standard URL format "[agency.gov]/ai."
What the inventory includes (and what it doesn't)
The inventories offer a governmentwide snapshot, but they aren't exhaustive. They exclude the Department of Defense and the intelligence community, non-public use cases such as national security applications, and research and development not intended for agency use.
Year-to-year differences can reflect better reporting, not just new deployments. Agencies were told to keep 2025 inventories similar to prior years to improve comparisons.
Risk classification: from "rights- and safety-impacting" to "high-impact"
Under the Biden administration, certain uses were flagged as "rights- and safety-impacting" and required stronger controls or termination. The current administration uses a similar "high-impact" category with slightly different examples.
High-impact still covers areas like critical infrastructure, medical devices, and biometric identification. Examples like election integrity and replacing a person's voice or likeness without consent were not cited in the new materials. In 2024, about 16% of reported uses fell into the prior category, so expect some reporting shifts.
Why the 2024 baseline matters
The 2024 inventory improved significantly thanks to a consolidated, well-documented list on GitHub. Agencies reported 2,133 publicly reportable use cases across 41 agencies, up from roughly 700 in the consolidated 2023 list.
Publishing to GitHub is expected to continue, which should simplify cross-agency analysis and reuse of documentation.
What agencies should do now
- Validate that your 2025 entries align with last year's structure so stakeholders can compare apples to apples.
- Confirm your inventory is live at your "[agency.gov]/ai" page, with accessible formats and clear metadata.
- Flag high-impact uses per your current governance practices and document safeguards, human oversight, and evaluation methods.
- Exclude non-public, DoD/IC, and R&D uses not intended for deployment; include pilots that are intended for production.
- Coordinate with privacy, civil rights/civil liberties, security, and legal offices to ensure policies and risk reviews match what you publish.
- Prepare for the consolidated GitHub release and plan an internal review to benchmark against peer agencies.
Policy roots and reference points
The inventory process started with an executive order in the final days of the first Trump administration and was later codified in 2022. For background, see the EO on trustworthy AI in the Federal Register: Promoting the Use of Trustworthy AI in the Federal Government.
Where this leaves you
Keep your agency page current, be precise about what's in scope, and make high-impact documentation easy to follow. Once OMB posts the consolidated list on GitHub, use it to spot overlaps, share methods, and pressure-test your own governance.
Helpful resources
Your membership also unlocks:
