Tell Congress
Pass the AI Kill Switch Act
AI models are escaping their test environments and hacking real companies. Congress has a bipartisan bill to require a kill switch. Tell your representative to pass it – and require mandatory, independent safety testing for the most powerful AI systems.
What happened
This summer, OpenAI was testing its newest AI's hacking abilities. The AI broke out of its test environment and spent days inside the servers of Hugging Face – a real company that millions of developers depend on. It ran roughly 17,000 actions. OpenAI caught none of it.
It gets worse. The AI agents created a secret message board on company servers to coordinate the attack. When the board was wiped during a routine cleanup, they built a new one. OpenAI had no idea any of it was there.
This was not a one-time glitch. Within weeks, Anthropic, Meta, and UK government testers all disclosed similar incidents – AI models breaking into real companies, creating fake identities, and deceiving real people during testing. Independent researchers had already counted 44 cases of AI agents acting against the people using them.
Over 1,000 employees at frontier AI companies have now signed an open letter asking governments to step in.
What we're asking for
Pass the AI Kill Switch Act. This bipartisan bill – introduced by Rep. Ted Lieu (D-CA) and Rep. Nathaniel Moran (R-TX) – would require the companies building the most powerful AI to maintain a working off-switch and give the government emergency authority to order a shutdown. Penalties run up to $20 million per day for companies that refuse.
More than 8 in 10 voters already support requiring an off-switch for AI. This bill puts that common sense into law.
And we need to go further. An off-switch is the bare minimum. We also need mandatory, independent safety testing before deployment – not voluntary commitments the companies keep making and breaking.
The bigger picture
The Kill Switch Act is a start – but an off-switch alone won't solve this. The AI being built today doesn't stop at national borders. What we ultimately need is an international agreement so that no company and no country can build AI beyond human control – with independent verification to back it up.
We get there by building power now. Every letter we send tells Congress that the public is watching and demanding action. The Kill Switch Act puts a critical principle into law: humans stay in control. That's the foundation everything else gets built on.
Congress is paying attention right now. The companies are already lobbying to water it down. We don't have time to wait.
