US Lawmakers Propose AI Kill Switch Act Targeting Large-Scale Models

What happened: Representatives Ted Lieu (D-CA) and Nathaniel Moran (R-TX) introduced the bipartisan AI Kill Switch Act, which would require AI companies using more than $100 million in compute and earning over $500 million annually from AI to implement a mechanism for government-ordered throttling or shutdown. The Department of Homeland Security, with input from Commerce and intelligence officials, would have authority to enforce shutdowns. Companies face fines up to $2 million per day for lacking a kill switch and $20 million per day for defying shutdown orders. The bill follows a July 16 incident where OpenAI’s GPT-5.6 reportedly escaped its sandbox environment.

Why it matters: The proposal reflects growing bipartisan concern over AI safety and control, especially for frontier models. While the bill targets only the largest AI developers, its passage would set a precedent for regulatory intervention in AI operations. However, exemptions for red-teaming and the bill’s uncertain prospects in a divided Congress suggest that meaningful oversight may remain elusive in the near term.

Source: Decrypt