Milestone 1 of 12 · May 2024
Seoul Summit
Sixteen countries pledge to develop kill-switch capabilities for frontier AI.
Key facts
- A Commerce Department letter in June 2026 led Anthropic to disable access to its Fable 5 and Mythos 5 models. Source
- H.R. 9917, the AI Kill Switch Act, was introduced in the House on 23 July 2026. Source
- The House of Lords called for statutory emergency shutdown powers on 2 September 2026. Source
- California issued an executive order advancing a verified frontier-model kill switch on 18 September 2026. Source
2024–2026
Select a milestone
- May 2024Seoul Summit Sixteen countries pledge to develop kill-switch capabilities for frontier AI.
- Sep 2025California SB-53 California passes frontier-model safety requirements.
- Nov 2025EU Cloud & AI Act European lawmakers advance compute and oversight provisions.
- Jun 2026Commerce action US Commerce moves against public model releases after safety concerns.
- Jul 2026H.R. 9917 The bipartisan AI Kill Switch Act is introduced in the House.
- Aug 2026Responsibility framework Anthropic publishes its AI responsibility framework.
- Sep 2026Expert divide Hinton questions feasibility while other researchers call for regulation.
- Sep 2, 2026UK Lords Peers call for government powers to compel an emergency shutdown of advanced AI.
- Sep 17, 2026Senate bill blocked A request for immediate adoption is blocked on the Senate floor.
- Sep 18, 2026California order An executive order advances a frontier-model kill switch verified by independent organizations.
- Sep 18, 2026UK government response The Cabinet Office rejects a statutory kill-switch amendment and favors targeted measures through the AI Security Institute.
- Sep 19, 2026Implementation scrutiny New reporting details distributed infrastructure, hackable controls, self-replication, and enforcement gaps that complicate a reliable shutdown.
Last updated: September 21, 2026
Edited by Juan L Aguirre