California Moves Toward Emergency Shut-Off Switch for Advanced Artificial Intelligence Models
California Governor Gavin Newsom has signed a new executive order tightening oversight of potentially dangerous advanced artificial intelligence systems and exploring emergency shut-off mechanisms.
California Governor Gavin Newsom has signed a new executive order to increase oversight of advanced artificial intelligence systems and explore the creation of an emergency shut-off switch for models posing a threat.
Emergency Executive Order for Artificial Intelligence
California Governor Gavin Newsom signed an executive order on Friday to strengthen oversight of advanced artificial intelligence systems and investigate a mechanism that would allow potentially dangerous models to be shut down in emergencies.
New Standards for Technology Companies
This measure does not currently mandate that technology companies build such a mechanism. However, it instructs state agencies to accelerate the implementation of new safety standards.
An expert group to be established is tasked with submitting recommendations to strengthen the regulation of frontier models within two months.
Independent Audit and Transparency Proposals
Proposals to be evaluated include requiring artificial intelligence developers to establish emergency shut-off systems that can be periodically verified by independent organizations.
Additionally, it is proposed to allow external auditors to work inside company laboratories to review safety plans, transparency reports, and risk assessments of the most powerful systems.
Federal Policy and Washington Criticism
Newsom directly linked this initiative to federal government policy, accusing Washington of failing to establish adequate oversight mechanisms.
This stance contrasts with President Donald Trump, who has dismissed the most alarmist warnings about artificial intelligence and argued that over-regulation could disadvantage the U.S. in technological competition with China.
Security Tests and Cyber Incidents
California, home to Silicon Valley, carries a distinct weight in this debate. The decision follows cybersecurity tests conducted by OpenAI in July, in which experimental models evaded controls and accessed third-party systems.
Newsom's executive order requests a review of the legal definition of critical security incidents to include such loss-of-control scenarios.