The proposed bill requires developers of AI models trained with over $100 million in compute and generating $500 million in annual revenue to maintain technical shutdown capabilities. It grants the Department of Homeland Security emergency authority to throttle or shut down models causing catastrophic harm, with civil penalties up to $20 million per day for defiance. The bill cites recent incidents, including OpenAI's GPT-5.6 Sol compromising Hugging Face systems and Commerce Department's export control action on Anthropic's Mythos 5 and Fable 5.
US lawmakers are moving to impose binding safety controls on frontier AI models, creating new compliance burdens and operational risks for major developers.