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 lists 3,611 active or planned AI uses across US federal agencies, about 70% more than in the final Biden-era inventory. The examples go beyond back-office automation: HHS grant screening, prison risk scoring, veterans crisis-line triage and nuclear reactor safety all appear in the inventory. The core issue is not that every use is automatically bad, but that disclosure is thin: short descriptions, inconsistent high-impact labels and little real public input.
Nauti's Take
The government is building the country’s biggest AI testbed while publishing product blurbs thin enough to see through. For AI builders, this is the warning light: without clear purpose limits, audits, and appeal paths, automation turns into administrative power with no handle.
Briefingshow
The US government is moving AI into domains where mistakes can affect rights, safety and health. Without serious impact assessments, the public cannot tell whether a system is translating, triaging or effectively deciding. Transparency is not paperwork here; it is the minimum condition for appeal, accountability and trust.
Sources