Newsom Orders AI Kill Switch and Independent Audits

California Governor Gavin Newsom has signed a new executive order aimed at accelerating independent oversight of artificial intelligence developers. The directive specifically calls for the creation of an emergency "kill switch" for AI models and responds directly to recent security vulnerabilities, such as the cyberattack targeting the Hugging Face platform. Under the order, an expert panel has been given a two-month deadline to deliver concrete regulatory recommendations.

Among the key proposals being evaluated is a requirement for AI companies to embed independent, third-party auditors directly within their research and development laboratories. Interestingly, this concept of internal oversight was originally suggested by several prominent AI labs themselves as part of a collective effort to safely manage the pace of frontier model development.

Newsom highlighted the current lack of federal oversight, pointing out that no national laws currently force AI companies to report dangerous safety incidents. He criticized political opponents, including Donald Trump, who have accused AI labs of fearmongering. For AI practitioners and developers, this executive order signals a rapid shift toward mandatory compliance and external validation. If California's framework is adopted nationally—as Newsom is urging Congress to do—developers will have to design models with built-in shutdown mechanisms and prepare for unprecedented levels of external scrutiny during the training phase.

The push for tighter controls comes amid growing anxiety from the scientific community. Recently, a group of 42 mathematicians issued a formal warning regarding the existential risks posed by advanced artificial intelligence systems, noting that these models could begin to evolve at an uncontrollable pace in the near future.