AI requirements are racking up across government, GAO says

Federal agencies are facing an onslaught of artificial intelligence requirements, a new government watchdog report detailed, with callouts coming from executive orders, federal laws, advisory guidance and other sources.

As of July, there were nearly 100 different objectives related to the emerging technology that might be considered government-wide standards, according to the Government Accountability Office. 

“AI technologies can drive economic growth and support scientific advancements that improve the conditions of our world,” the GAO said in its correspondence to Congress. “It also holds substantial promise for improving the operations of government agencies. However, AI technologies also pose risks that can negatively impact individuals, groups, organizations, communities, society, and the environment.”

The goal of the report, the watchdog said, was to understand the various AI requirements facing the government and which bodies hold responsibility related to the technology. 

The review included current requirements for federal agencies, like creating inventories of AI use cases and updating AI use policies. It also examined broader efforts, like the National AI Initiative, which focuses on goals like increasing research and development of the technology and investing in computing resources. 

“Federal agencies’ efforts to implement AI have been guided by a variety of legislative and executive actions, as well as federal guidance,” GAO continued. “Congress has enacted legislation, and the President has issued EOs, to assist agencies in implementing AI in the federal government.”

The office reviewed new artificial intelligence initiatives created by the current and former administrations, stretching from the first Trump administration’s executive order on artificial intelligence and the signing of the AI Training Act to more recent guidance from the Office of Management and Budget.

Overall, GAO found that 10 different bodies had a stake in reviewing the U.S. government’s AI efforts, and that federal laws, executive orders, and guidance had produced 94 different expectations related to the technology, including reviews related to risk mitigation, investment strategies, and usage policies. 

The GAO sent a draft of the report to OMB, the Office of Science and Technology Policy, the Commerce Department, the General Services Administration and the National Science Foundation. OSTP, Commerce and NSF responded with technical comments, while GSA declined to provide comments and OMB did not respond to GAO’s request for comments on its findings.