Governments may use industrial‑scale emergency authorities (like the U.S. Defense Production Act) to force frontier AI companies to produce models the military can use for any lawful purpose, even if firms had contractually restricted certain uses. That dynamic turns safety or ethics guarantees into bargaining chips that can invite legal coercion, supply‑chain blacklisting, or forced nationalization of AI capabilities.
— If adopted more broadly, this approach would remake AI governance: safety concessions could be reversed by state power, chilling private safety commitments and concentrating control of frontier systems in the state.
Scott
2026.03.10
90% relevant
Aaronson reports that the Pentagon threatened to invoke the Defense Production Act (DPA) to commandeer Anthropic’s technology and describes the simultaneous use of a 'supply chain risk' label—precisely the kind of state coercion (compelling military use or transfer of private AI IP) encapsulated by the existing idea.
Kelsey Piper
2026.02.26
100% relevant
Anthropic’s contract stipulations (no autonomous killing, no domestic mass surveillance) and the DOD threats reported in Semafor/Axios — including an ultimatum from Defense leadership and plans to invoke the Defense Production Act or declare a supply‑chain risk.
← Back to All Ideas