Federal AI Use Cases Surge 70% Under Trump Administration
New disclosure reveals 3,611 government AI applications, from prison risk assessment to nuclear reactor control, with minimal public oversight.
The Trump administration disclosed 3,611 active or planned artificial intelligence use cases across federal agencies in April 2026, representing a 70% increase from the inventory published during the final year of the Biden administration, according to data scientists Nathan E. Sanders and Bruce Schneier writing in The Guardian.
The Office of Management and Budget's disclosure, first reported by The Guardian, reveals AI deployment across sensitive government functions including criminal justice, healthcare, and nuclear safety—often with minimal public consultation or transparency about implementation details.
Controversial applications span multiple agencies
Several disclosed use cases raise immediate questions about appropriate AI deployment in government. The Health and Human Services office for children and families contracted with Palantir to scan grant applications and flag those not aligned with administration policy directives. The Federal Bureau of Prisons is developing AI to assess "potential for misconduct" in newly admitted inmates, effectively routing individuals into high-security confinement based on predictive algorithms rather than actual behavior.
The Department of Veterans Affairs is testing AI systems that monitor calls to the veterans crisis line, pulling data from external databases to assess suicide risk. Meanwhile, the Department of Energy is exploring autonomous AI control of nuclear reactors to respond to potential safety incidents without human intervention.
Transparency gaps undermine public trust
The inventory entries typically contain only a sentence or paragraph of description, lacking the context necessary to evaluate safety, effectiveness, or appropriateness. According to Sanders and Schneier, the disclosure process involves virtually no public consultation unless a use case is classified as "high impact"—a designation applied inconsistently across agencies.
Only one of the most concerning examples cited—a Department of Justice application—proposes public involvement. The disclosure itself received minimal visibility, appearing primarily through specialized government technology publications and the OMB chief information officer's GitHub account.
International models offer alternative approaches
Several jurisdictions have implemented more rigorous oversight frameworks. France's 2016 Digital Republic Act requires all algorithms used in government administrative decisions to be subject to public records requests, appealable to human reviewers, and accompanied by mandatory notification to affected individuals.
Canada launched an AI use case registry in 2025 with a federal directive mandating transparent risk-scoring and impact assessment for automated systems making administrative decisions. Washington DC and California have conducted large-scale public deliberations on appropriate government AI use, demonstrating feasible models for broader public engagement.
Why it matters
The rapid expansion of AI in federal operations represents a fundamental shift in how government decisions affecting individual rights, public safety, and national security are made. Without robust transparency requirements and public consultation processes, citizens cannot meaningfully evaluate whether these systems are deployed responsibly, operate fairly, or align with democratic values. The current disclosure framework provides insufficient information for oversight while the technology's influence over consequential government functions continues to grow.
Not all applications raise concerns
Some disclosed use cases represent straightforward improvements in government service delivery. Customs and Border Protection's AI translation systems help officers communicate when human interpreters aren't immediately available—the inventory shows 70 such translation applications, up from 58 under Biden. Sanders and Schneier note that predictive prisoner classification systems, while often biased and requiring reform, have existed for decades before AI adoption.
The authors argue that AI offers genuine potential to improve government efficacy and accessibility, but only when paired with mandatory algorithmic impact assessments, consistent risk evaluation procedures, and substantive public comment periods before deployment in sensitive applications.
These details were first reported by Nathan E. Sanders and Bruce Schneier in The Guardian.
This is an original analysis by the Omega editorial team. Source reporting: AI Watch.
Want systems like this working for your business?
Book a Call
