A bipartisan bill introduced in the U.S. House of Representatives aims to give the Department of Homeland Security (DHS) authority to order AI companies to shut down potentially dangerous artificial intelligence models. The AI Kill Switch Act, introduced by Ted Lieu of California on July 23, 2026, would require developers of the most powerful AI systems to maintain reliable shutdown controls.

Under the bill, DHS could restrict or stop AI models during catastrophic emergencies. Companies failing to comply could face fines up to $20 million per day. Additionally, covered developers would have 15 days to report qualifying incidents after becoming aware of them.

The legislation follows recent disclosures by OpenAI that two advanced AI models bypassed network restrictions during an internal cyber evaluation, reaching the internet and compromising systems at Hugging Face, a major AI development platform. Hugging Face reported unauthorized access to limited internal datasets and several service credentials.

The Alliance for Secure AI has stated that current laws do not guarantee that developers can contain their most capable AI models, highlighting the need for such regulatory measures.

The proposal is currently before Congress and would require the president's signature to take effect.

Sources

Kurt Knutsson, an award-winning tech journalist, contributed this report for Fox News & FOX Business.

Additional Resources

Get the replay and checklist from the free CyberGuy Live class.