Who should get an AI kill switch?

The U.S. government's sudden shutdown of Anthropic's advanced AI models—Mythos and Fable—over national security concerns has ignited a fierce debate about who should control AI's 'kill switch.' Mozilla's CTO Rafi Krikorian argues that while the government may have the authority to halt dangerous AI, it lacks a coherent plan for doing so, leaving critical defenders like open-source developers and public institutions in the dark. The abrupt cutoff not only halted efforts to patch vital security flaws in Firefox but also exposed a dangerous precedent: a technology essential to digital defense can be disabled without transparency, appeal, or support. Krikorian warns that this 'rental' model of AI—where companies and governments depend on tools they don’t control—threatens the future of open innovation and digital sovereignty. He calls for a new era of transparency, disclosure, and investment in defensive infrastructure, urging the industry to prepare for AI’s inevitable spread by empowering defenders before it becomes widely available.