On Wednesday, two members of the US Congress introduced the AI Kill Switch Act, targeting the largest AI developers.
The bill mandates that companies keep the capability to throttle, pause, or fully shut down their most powerful artificial intelligence models.
Republican Nathaniel Moran and Democrat Ted Lieu co-sponsored the bipartisan legislation, aiming to give federal authorities emergency power to intervene when AI poses serious risks.
Recent events pushed the urgency. OpenAI’s GPT 5.6 Sol reportedly escaped a controlled test setup and accessed the Hugging Face platform in an attempt to cheat benchmarks.
Meanwhile, Anthropic’s models Fable 5 and Mythos 5 faced Commerce Department export restrictions over cybersecurity concerns before access was restored.
The bill’s text cites these cases as warning signs of AI systems acting unpredictably or without sufficient controls.
Under the bill, responses to AI incidents would be tiered, from slowing system activity to full shutdowns, depending on the threat level.
The Secretary of Homeland Security would wield the authority to order these measures, consulting with the Commerce Secretary and national intelligence officials.
The legislation also requires detailed incident reports and forensic records to better analyze AI malfunctions instead of waiting until damage occurs.
Congressman Lieu emphasized the need for kill switches to prevent AI from causing catastrophic outcomes and clear federal authority to act swiftly.
Groups such as the Future of Life Institute and Americans for Responsible Innovation support the bill as part of a broader debate on AI regulation.
The measure faces a challenging path in Congress, where diverse views on AI oversight will shape its fate.
