AI use by the US government is ballooning. And the lack of transparency is troubling | Nathan E Sanders and Bruce Schneier
TL;DR
The OMB disclosed on April 14 that US federal agencies have 3,611 active or planned AI use cases, about 70% more than the final Biden-era inventory, according to Nathan E Sanders and Bruce Schneier. Examples include HHS grant screening with Palantir, prison misconduct risk scoring for new inmates, AI support for Veterans Crisis Line calls, and Department of Energy tests around nuclear reactor safety.
Nauti's Take
This is not a government demo stack; it is decision machinery for grants, prisons, crisis lines, and nuclear safety. If you build AI for public agencies, audit trails, appeal paths, and readable registries are product requirements, not paperwork stapled on after launch.
Briefingshow
When agencies use AI in areas like liberty, health, safety, and public benefits, a raw inventory is not enough. The real test is whether affected people know automation is involved, can challenge decisions, and can see risks assessed before deployment. Without that layer, efficiency becomes a quiet transfer of power.
Sources