Governor Gavin Newsom issued an executive order on Friday directing California officials to establish a task force that will propose regulations for artificial intelligence safety within two months. The directive specifically asks the group to consider mandating a kill switch for frontier models, alongside requirements for independent onsite audits and verified transparency reports. This move positions the state to lead national efforts on AI oversight by introducing concrete mechanisms for external verification of high-risk systems. The resulting recommendations will inform how California law addresses the potential for uncontrolled deployment of advanced models.
The practical impact lies in shifting accountability from voluntary corporate promises to enforceable state standards. By requiring independent auditors to verify risk assessments, the state aims to prevent companies from claiming safety without third-party confirmation. A mandated kill switch would provide a concrete method to halt system operation if specific safety thresholds are breached, rather than relying on internal company protocols. This approach prioritises external checks over self-regulation in a sector where failure rates remain high.
