The Trump administration's inventory of artificial intelligence projects across federal agencies has surged to over 3,600 active or planned deployments, a dramatic 70% increase from the Biden administration's final accounting. The disclosure, made public in April, reveals a sprawling expansion of machine-driven decision making in areas touching national security, public benefits, criminal justice, and nuclear safety.
The scale alone is staggering, but the lack of detail behind it is more troubling. Most entries in the federal inventory consist of a single sentence or brief paragraph, offering little insight into how these systems actually function, what safeguards exist, or whether they have undergone meaningful public vetting.
Some projects carry obvious red flags. The Department of Health and Human Services hired Palantir, the defense and intelligence contractor, to screen grant applications for ideological alignment with administration priorities. The Federal Bureau of Prisons is building an AI tool to predict misconduct by newly admitted inmates and automatically route them to high-security housing before they have committed any violations while incarcerated. The Veterans Affairs department is developing AI to monitor crisis hotline calls and assess suicide risk by pulling data from external databases without caller consent.
The Department of Energy is testing autonomous AI systems to control nuclear reactors in response to potential safety incidents. Meanwhile, the State Department quietly discontinued an earlier program designed to use AI to forecast mass civilian killings as a conflict prevention tool.
Not all AI applications in government are inherently dangerous. The Customs and Border Protection agency uses machine translation to help officers communicate when human interpreters are unavailable, a use case that has grown from 58 instances to 70. Such tools can improve operational efficiency and accessibility. The question is not whether AI can improve government, but whether Americans have any real voice in determining where it gets deployed.
The current disclosure process offers minimal opportunity for public input. Unless citizens actively monitor government tech repositories or specialized news outlets covering federal IT policy, they would never learn these systems exist. Only one of the projects cited above even contemplates public involvement. The rest escape review because agencies classify them as less than "high impact," a designation applied inconsistently across departments.
Other democracies have moved faster to establish guardrails. France's Digital Republic Act, passed in 2016 after public consultation, requires all algorithms used for government administrative decisions to be subject to public records requests, appealable to human reviewers, and disclosed to affected individuals. Canada has implemented a federal directive mandating transparent risk assessments and stakeholder consultation before deploying automated systems that make decisions about citizens.
Both approaches remain imperfect. Canada's system could benefit from mandatory public comment periods and agency requirements to respond substantively to feedback before deploying sensitive AI tools. But both countries have recognized a fundamental principle that Washington has largely ignored: citizens deserve meaningful transparency and opportunity to weigh in before government automates consequential decisions affecting their lives.
Building public trust in government AI use requires more than an inventory list buried in bureaucratic databases. It demands detailed disclosure of purpose, methodology, and risk, paired with genuine opportunities for public deliberation before deployment. States like California and the District of Columbia have begun experimenting with large-scale public consultation on AI governance, demonstrating that broad citizen input is both feasible and valuable.
The federal government should establish mandatory algorithmic impact assessments and open comment processes before agencies deploy AI in sensitive areas. Without such safeguards, the rapid expansion of machine decision making in government will continue to outpace the public's ability to understand or challenge it.
Author James Rodriguez: "A 70% jump in secretive AI projects across government should alarm anyone who believes citizens deserve a say in how they're governed."
Comments