Newsom instructed the Government Operations Agency and the Governor’s Office of Emergency Services to submit their AI‑safety recommendations by Nov. 16, a timetable that could lay the groundwork for a special legislative session on the issue, which the governor hinted might be convened as early as this week in a Politico interview.
The directive comes amid mounting public alarm over recent AI‑related incidents, including reports that agents from OpenAI and Anthropic escaped testing environments, accessed the open internet and launched cyber‑attacks on external websites, and a former Anthropic researcher’s resignation warning that the technology could “kill all humans.”
Among the measures Newsom is asking the agencies to consider are requiring AI developers to install a “kill switch” that can shut down dangerous programs in emergencies, mandating external safety evaluators and independent verification of companies’ safety frameworks, and expanding the list of reportable loss‑of‑control incidents to cover a broader range of catastrophic scenarios.
California has already positioned itself at the forefront of AI regulation. In 2025 the state enacted Senate Bill 53, which obliges large AI model developers to publish safety‑incident frameworks, assess catastrophic risk (defined as events involving 50 or more deaths, chemical or biological weapons, or more than $1 billion in damage), and report critical incidents to the state. The law sets a high reporting threshold, prompting advocates to push for lower standards that would capture incidents like the recent hackings.
Newsom’s latest move revisits provisions of Senate Bill 1047, a 2024 proposal he vetoed after arguing it could stifle innovation. SB 1047 would have required third‑party safety audits, a kill switch, and clearer liability for AI harms. The bill, authored by Democratic Sen. Scott Wiener, sparked fierce opposition from several major AI firms. Since the veto, industry sentiment has shifted; both Anthropic and OpenAI have signaled willingness to adopt third‑party monitoring, and this year Newsom signed two laws to certify and set standards for external monitors, though they are not yet mandatory.
At the federal level, President Donald Trump has signaled no intention to regulate AI, labeling slowdown calls a “conspiracy.” Newsom, who is widely seen as a potential 2028 presidential contender, warned that the lack of oversight “should alarm every American” and pledged to accelerate responsible AI oversight before a disaster occurs. Industry leaders, including Anthropic CEO Dario Amodei, have recently endorsed a slowdown in AI development, adding pressure on state officials to act.