The Rapid Rise of Federal AI Use Demands Greater Transparency and Public Oversight

Text
The Trump administration disclosed 3,611 active or planned artificial intelligence (AI) use cases across federal agencies as of April 14, 2025, according to the Office of Management and Budget (OMB). This represents a 70% increase from the 2,124 cases reported in the final year of the Biden administration, highlighting a rapid expansion of AI integration into government operations. The inventory, first revealed in a federal chief information officer’s GitHub update, includes applications ranging from automated decision-making in healthcare to nuclear reactor management, raising concerns about transparency and accountability.

Subheading
AI in Sensitive Government Functions
The OMB’s inventory details AI systems handling tasks with significant societal impact, including evaluating inmate misconduct risk, assessing suicide risk for veterans, and monitoring nuclear reactor safety. For example, the Department of Health and Human Services (HHS) partnered with Palantir, a company known for its work with military and law enforcement agencies, to analyze grant applications for ideological alignment with administration policies. Similarly, the Federal Bureau of Prisons is developing an AI tool to predict misconduct among new inmates, potentially leading to preemptive high-security placements.

Text
The Department of Veterans Affairs (VA) is testing AI to evaluate suicide risk during crisis line calls by analyzing voice patterns and cross-referencing external data. The Department of Energy (DOE) is also exploring AI for autonomous nuclear reactor control, a practice that critics argue could prioritize efficiency over human oversight. Meanwhile, the State Department recently terminated an AI-driven program to forecast mass civilian killings, a decision described as “disturbing for its retirement rather than its deployment” in the source material.

Subheading
Transparency Gaps and Public Consultation Challenges
Despite the scale of AI adoption, the OMB’s inventory provides minimal context, with most entries offering only brief descriptions. Only one cited example—the Department of Justice’s (DOJ) public consultation process—meets the administration’s “high impact” criteria, which are applied inconsistently across agencies. According to the source, the lack of public engagement means most citizens remain unaware of these developments unless they follow specialized outlets like FedScoop or monitor federal agency repositories.

Text
The authors of the source material, Lisa Park and Nathan E. Sanders, note that while some AI applications, such as machine translation in Customs and Border Protection (CBP), could improve efficiency, others risk exacerbating biases or eroding civil liberties. For instance, predictive algorithms used to classify prison risk have historically been prone to racial and socioeconomic bias, despite their long-standing use. Similarly, AI-driven nuclear reactor management, though technically established, raises questions about accountability in the event of autonomous system failures.

Subheading
Comparative Approaches to AI Governance
France and Canada offer contrasting models for AI transparency. France’s 2016 Digital Republic Act mandates that automated government decisions be subject to public records requests, appealable by humans, and accompanied by notification to affected individuals. Canada’s 2025 AI use case registry includes a federal directive requiring risk assessments and stakeholder consultations for automated systems. While these frameworks are not perfect, they illustrate how structured oversight can balance innovation with accountability.

Text
The U.S. lacks a centralized policy for AI transparency, with the OMB’s inventory serving as the primary public record. Critics argue that the current approach risks entrenching opaque systems without public input. The source material emphasizes that AI’s potential to enhance government efficiency—such as CBP’s 70 AI translation use cases, up from 58 in 2024—must be weighed against the need for rigorous oversight.

Subheading
Calls for Reform and Public Engagement
The authors advocate for algorithmic impact assessments and public comment periods to address gaps in U.S. AI governance. Washington D.C. and California have initiated public deliberations on AI use, leveraging digital platforms to gather input. These efforts align with broader demands for democratic accountability in an era of rapidly advancing technology.

Text
The source material, originally published in The Guardian and co-authored by Park and Sanders, underscores that AI’s role in government is neither inherently beneficial nor harmful. Its impact depends on implementation, oversight, and public trust. As the OMB’s inventory grows, the challenge lies in ensuring that AI systems serve the public interest without sacrificing transparency or equity.

Quoted text
“AI offers real potential to improve the efficacy, efficiency, and accessibility of government. But, equally, there is legitimate reason for public concern and distrust that can only be addressed through transparency and dialog.”
Source: Lisa Park and Nathan E. Sanders, The Guardian