California Orders AI Kill-Switch Review as Lawmakers and Labs Confront Shutdown Problem

California has moved to accelerate oversight of frontier AI systems and examine whether advanced models should be required to include a verifiable emergency shutoff. The push comes as recent AI-agent security incidents have turned a once-theoretical safety question into an operational one: if a powerful AI system behaves dangerously, who can stop it, how fast, and with what proof that the stop actually worked?