The U.S. government is going all-in on AI without a plan
The U.S. Government Accountability Office (GAO) just warned that federal agencies are rapidly adopting AI across hundreds of use cases without unified safety and procurement standards. While agencies already use AI in over 200 operations, oversight officials found that regulatory guidelines are years behind schedule. Why it matters: Rushing deployment without shared rules leaves sensitive tools like facial recognition wide open to security gaps and civil liberties abuses.
The federal government wants Silicon Valley to slow down and follow the rules on AI. But inside its own agencies, Uncle Sam has been adopting the technology at breakneck speed — with almost no guardrails in place.
A new 96-page report from the Government Accountability Office (GAO) just pulled back the curtain on how Washington actually uses artificial intelligence. The watchdog surveyed 23 non-military agencies and found federal departments are years behind on establishing standard rules to responsibly buy and deploy AI tools.
Hundreds of secret use cases
According to the report, federal agencies currently run 228 active AI projects, with nearly half launched in the past year alone. If you count planned projects, the total jumps to 1,241 applications across government.
About 70% of these use cases look harmless enough on paper. NASA uses machine learning to monitor volcanic activity worldwide, while the Commerce Department deploys computer vision to track wildfires and count seals, walruses, and seabirds from drone imagery.
However, transparency drops off a cliff when it comes to sensitive operations. Federal agencies refused to publicly disclose more than 350 active and planned AI tools. The State Department, for instance, listed 71 distinct AI use cases but agreed to make only 10 public.
Civil liberties groups are particularly alarmed by law enforcement deployments. The Department of Justice and the Department of Homeland Security reported only 25 total use cases between them, but those include border surveillance cameras and facial recognition. A previous GAO finding revealed that between 2019 and 2022, 95% of thousands of facial recognition searches across six agencies were run without proper staff training.
Two years behind schedule
The bigger problem is bureaucratic paralysis at the top. Under a 2020 law, the Office of Management and Budget (OMB) was required to issue government-wide guidelines for acquiring and using AI by September 2021.
OMB blew past that deadline by more than two years, only releasing a draft memo last month following President Biden’s comprehensive AI executive order.
The void has left agencies flying blind. The Office of Personnel Management failed to deliver a required forecast of federal AI workforce needs for the next five years. Meanwhile, 10 major departments, including Treasury and Education, still have no concrete system to track or update their AI inventories over time.
Why it matters
Washington is projecting strength on AI safety abroad and in executive orders, but its domestic house is a mess. Demanding that frontier AI labs undergo independent red-teaming carries far less weight when federal law enforcement is already deploying commercial facial recognition and location data without basic training standards.
For enterprise tech vendors, this is a clear warning sign. The federal government is eager to procure commercial machine learning tools, but the lack of unified guardrails will inevitably lead to botched deployments and public backlash. Startups aiming to sell algorithmic tools to public agencies should expect procurement freezes and intense congressional scrutiny as Washington scrambles to clean up its own blind spots.
Sources (1)
- cnn.com(opens in a new tab)Originalcnn.com