U.S. Congress bill would require a kill switch for AI systems trained on $100M+
rohanpaul_ai · x · 2026-07-25
Congress bill would force frontier AI systems to have a kill switch
A bipartisan House bill would require a working kill switch for AI systems that cost $100M+ to train. Enforcement would sit with DHS, because the proposal amends the Homeland Security Act.
- Covered systems start at $100M training compute and $500M annual revenue.
- Developers must be able to stop inference, cut user access, and shut the system down.
- DHS could also order throttling, disabling, or full shutdown after consulting Commerce and the intelligence director.
- Sabotaging a shutdown command would be punishable; incidents causing 10 deaths or $100M damage trigger reporting.
- Companies would have to report within 15 days and preserve weights and telemetry after an order.
- Fines could reach $2M/day, rising to $20M/day for defying an emergency order.
The post also notes a core limitation: hosted models can be stopped, but downloaded weights may not obey a kill switch at all.
Related event: US House Proposes AI Kill Switch Act Amid OpenAI Security Incidents(7 posts)→
More from AGI Musings
- Physicists Let AI Pick Its Own Problem—It Wrote a Publishable Paper in 2 Hours — burny_tech · 2026-10-27
- Deep learning veteran warns: don't let entrenched disbelief derail prep for the next big tech — jachiam0 · 2026-10-27
- Should platforms serve third-party AI agents? The antitrust question — sjgadler · 2026-10-27
- Is an agent clicking 'I am human' a lie? Debate over Amazon blocking agents — sjgadler · 2026-10-27
- Musk: robot RAM demand will dwarf everything else, and 100% of superintelligence compute goes to deep space — elonmusk · 2026-10-27
- Anthropic S-1 warns of AI job displacement as Microsoft offers exit to ~27% of staff — _jaydeepkarale · 2026-10-27