AI use by the US government is ballooning. And the lack of transparency is troubling | Nathan E Sanders and Bruce Schneier
TL;DR
The US federal government now lists 3,611 active or planned AI use cases, about 70 percent more than in the final Biden-era inventory, according to The Guardian. Examples include Palantir scans of HHS grant applications, risk scoring for newly admitted prisoners, AI on the Veterans Crisis Line, and tests for AI responses to nuclear safety incidents.
Nauti's Take
This is the dangerous middle phase: enough AI to shift real power, not enough transparency to pin down responsibility. A registry is only the starting point.
Sensitive use cases need risk assessments, public comment, appeal paths, and a human reviewer who is more than a rubber stamp. Otherwise, public-sector modernization turns into a black box with an agency logo.
Briefingshow
Government AI is not just an efficiency story because it can enter decisions about liberty, benefits, crisis support, and safety. If agencies publish only inventory lines, citizens, courts, and oversight bodies cannot tell whether a system merely assists, screens people out, or effectively decides.
Sources