The US government’s annual tally of AI uses is on its way
An inventory documenting federal uses of artificial intelligence in 2025 is being currently prepared for publication by the White House Office of Management and Budget, according to the Trump administration’s top IT official.
In a statement to FedScoop, 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.”
The anticipated release will mark the latest reporting year for the annual public AI use case disclosures and the first during the Trump administration, which has pushed for expanded use of the technology to fuel its efficiency aims.
News that the inventories are imminent comes after deadlines circulated internally this summer already passed. According to June internal guidance obtained by FedScoop, the administration set a deadline of Nov. 4 for agencies to submit their inventories to OMB and a deadline of Dec. 2 for officials to post that information on agency websites. That initial deadline, however, fell within the government shutdown.
During a December interview with The Daily Scoop Podcast, Barbaccia acknowledged that the administration was taking the shutdown days and the then-upcoming holiday season into account when setting expectations for those inventories, but noted that it wouldn’t be “a major delay.”
The federal AI use case inventory process was the result of an executive order inked in the final days of the first Trump administration. It then began during the Biden administration and was codified into statute in 2022.
While the inventories were originally intended to improve transparency, help agencies coordinate uses, and inform policy guidance on use of the technology, it’s taken a few years for the publications to become useful. Early iterations of those inventories were rife with inconsistency, delays, and sometimes even errors. The process generally improved with the 2024 publication following Biden administration efforts to “enhance” the process.
That inventory ultimately included a total of 2,133 publicly reportable use cases across 41 federal agencies. A consolidated list put out in 2023, by comparison, reported roughly 700 use cases across inventories.
Changes to the 2024 collection included publishing a consolidated list of federal AI use cases on the development platform GitHub, which allowed for easier access to governmentwide data and better documentation. Barbaccia’s statement indicates publishing to that platform will continue.
The inventories serve as a barometer for the government’s adoption of the technology, though it’s important to note that they aren’t designed to include every single use of AI across the federal government.
The process doesn’t include the Department of Defense or elements of the intelligence community. The disclosures also don’t encompass non-public use cases — such as those being used for national security — or research and development with the technology that isn’t intended to ultimately be used by agencies.
What should or shouldn’t be included, such as R&D, has been a point of confusion for agencies in the past. As a result, some year-to-year changes in the inventories are purely attributable to improved reporting over time.
Per the June guidance obtained by FedScoop, inventories are expected to be relatively similar to previous years, which provides a better chance for improved year-to-year comparison. The biggest part of the process that has changed is how agencies handle uses that require greater risk management.
The Biden administration introduced a classification known as “rights- and safety-impacting” uses that must follow more stringent management practices or be terminated. The Trump administration has a similar category in its governance practices called “high-impact” uses, but it differs slightly.
Similarly to the rights- and safety-impacting classification, whether a use is high-impact still must be disclosed on the inventory, per the guidance obtained by FedScoop. High-impact uses, however, cover a slightly different group of applications. While critical infrastructure uses, medical devices, and biometric identification all still fall under the new high-impact umbrella, use cases for election integrity and replacing a person’s voice or likeness without consent weren’t mentioned in the new examples.
It’s not yet clear how those differences will change reporting in the inventory, but roughly 16% of uses reported in 2024 fell into the previous category.
Another aspect of the process that’s different from previous years is the absence of publicly available guidance to agencies on the inventories. In the past, the guidance has been published online and sets public expectations for deadlines and reporting criteria. FedScoop obtained that guidance independently.
So far, only a handful of agencies appear to have documented their 2025 inventories on their websites, such as the Office of Personnel Management and the Nuclear Regulatory Commission. Per the guidance, when the inventories are public, they should be on each agency’s website at “[agency.gov]/ai.”
