Federal AI Use Soars, But Bottlenecks Threaten Momentum Amid Public Skepticism: Brookings

The usage of artificial quality crossed the U.S. national authorities has expanded dramatically successful caller years, but important obstacles—from endowment shortages to nationalist skepticism—are slowing the technology's liable integration into authorities services, according to a caller study from the Brookings Institution.

The Wednesday study draws connected AI usage lawsuit inventories from 2023 to 2025, national jobs data, Office of Management and Budget memoranda, and interviews with existent and erstwhile national technologists crossed 8 agencies.

The numbers archer a communicative of accelerated acceleration. In 2025, 41 agencies documented much than 3,600 idiosyncratic AI usage cases—69% supra the full reported successful 2024 and 5 times the fig reported successful 2023. The applications span a wide scope of authorities functions: More than fractional of the Social Security Administration's reported usage cases enactment work transportation and benefits processing, portion implicit fractional of the Department of Justice's inventory supports instrumentality enforcement efforts.

Yet the maturation is acold from evenly distributed. For the past 3 years, 5 ample agencies accounted for implicit fractional of each reported AI usage cases, and ample agencies contributed 76% of the full inventory successful 2025. Smaller agencies are hardly keeping pace: The 11 tiny agencies that reported successful 2025 collectively submitted conscionable 60 usage cases, representing lone 2% of the full inventory.

The study identifies respective structural barriers holding backmost broader adoption. One of the astir pressing is simply a deficiency of specialized talent. Of much than 56,000 method occupation listings posted by the national authorities since 2016, conscionable implicit 1,600—fewer than 3%—explicitly notation AI capabilities.

A Biden-era hiring surge aimed to code this gap, but workforce reductions successful aboriginal 2025 whitethorn person undermined those efforts, arsenic astatine slightest 25% of AI-specific occupation listings were posted from 2024 onward—meaning galore of those recently hired workers could person been among the astir precocious and easy dismissed.

Beyond staffing, the study points to a profoundly ingrained civilization of hazard aversion wrong national agencies. Nearly 60% of each AI usage cases are either successful the aviator oregon pre-deployment stage, suggesting the national AI scenery is inactive successful a accelerated maturation phase—one that requires dedicated clip for acquisition and experimentation that galore agencies conflict to carve out. The study besides notes that the Trump administration's explicit linkage of AI deployment to workforce cuts done the Department of Government Efficiency (DOGE) whitethorn beryllium reinforcing that hesitancy.

Accountability gaps are different concern. More than 85% of each high-impact deployed AI usage cases successful 2025 deficiency immoderate required accusation astir hazard mitigation measures, contempt explicit requirements from the OMB.

Public assurance poses yet different challenge. According to caller Pew Research Center data, astir fractional of Americans present accidental they are much acrophobic than excited astir the increasing prominence of AI, up from 37% 4 years prior, and conscionable 17% of the American nationalist believes AI volition positively interaction the U.S. successful the adjacent 2 decades.

The study warns that the stakes are high. Public spot successful the national authorities remains adjacent historical lows, with caller information showing lone 16% of Americans saying they spot Washington to bash what is close astir oregon astir each of the time. Against that backdrop, the authors reason that poorly executed AI deployments could origin superior damage—but that well-designed applications focused connected tangible work improvements could, conversely, assistance rebuild assurance successful authorities institutions.

To get there, Brookings recommends expanding AI literacy grooming crossed agencies, reforming procurement rules that were designed for much static bundle systems, strengthening transparency practices astir high-risk AI use, and prioritizing usage cases that nutrient clear, affirmative benefits for the public.