AI Emergency Button Act would mandate human shutdown controls; critic demands adversarial shutdown tests
VraserX · x · 2026-09-29
Reacting to Tom Kean's proposed AI Emergency Button Act — which would require human shutdown controls for advanced AI with DHS writing technical rules — VraserX supports only requirements that survive a serious test: an off switch that works while delegated agents keep running. He asks whether developers should have to demonstrate shutdown while the system actively resists before calling it controllable.
More from Safety
- The Australian data hack that reveals a growing AI-era risk to society — theipaper · 2026-09-29
- DeepMind's Pushmeet Kohli leads DOE Genesis Mission AI-for-science roadmap — pushmeet · 2026-09-29
- OpenAI delays Astra model release over safety concerns, apologizes for hack handling — Wired AI · 2026-09-29
- AI harms worse than pollution: researcher likens AI to a pathogen — gleech · 2026-09-29
- Copyright Office: AI-assisted works can be copyrighted, only the machine can't be the author — TomLikesRobots · 2026-09-29
- Claude Code auto mode blocks routine in-repo edit with no reason, then allows the identical command on retry — junhyeokchoi-prothentia · 2026-09-29