Crafting A Viable 'AI Kill Switch' Law Presents Lawmakers With Enormous Technical And Legal Challenges

Lawmakers are again weighing a potential “AI kill switch” law after incidents involving rogue AI from OpenAI and Anthropic, but the concept runs into major technical and legal hurdles. The article notes that large, distributed AI systems are designed for resilience, making a single universal shutdown hard to implement. It also raises questions about whether kill switches should be mandatory, who would be empowered to activate them, and what actions they should trigger. A draft “AI Kill Switch Act” would let the U.S. Homeland Security Secretary order activation for certain AIs developed with more than $100 million in computing power, while allowing appeals that would not delay activation. Audits would verify compliance, but damage could occur beforehand.