California Gov. Gavin Newsom released recommendations November 18 for mandatory AI safety standards, independent oversight and emergency shutdown powers, his office announced.

Why it matters: The proposals target large developers, not small-business users. Owners can map AI-dependent customer workflows, check vendor continuity information and identify manual or alternate-tool fallbacks without treating this as new law.

Catch up quick

  • Anadolu identifies the November 16 report, “AI Safety in 2026 and Beyond,” as prepared by the California Government Operations Agency, consulting the California Office of Emergency Services and outside experts.
  • Newsom’s office cites two bills signed in 2026: SB 813 establishes independent-verification frameworks; AB 1405 creates an AI-auditor registry and independence standards.
  • Anadolu reports: “The proposals are not legally binding and would require legislative action” and “California currently has no immediate shutdown authority under the proposed framework”.
  • California Attorney General Rob Bonta’s OpenAI safety lawsuit alleges inadequate development/testing safeguards and monitoring, the Los Angeles Times reports.
  • Newsom recommends independent verification across three stages: training, development and deployment. Anadolu says licensed organizations would translate standards into tests, receive sufficient model access and submit unredacted findings to regulators.

What’s actually new

  • CalMatters says shutdown powers address developers refusing containment. Anadolu identifies triggers: credible evidence of imminent catastrophic threat or serious risk of losing human control.
  • Anadolu says orders would apply immediately, pending judicial review; developers could seek temporary restraining orders. Responses must be least restrictive but effective, ending when imminent danger passes.
  • Newsom’s office proposes verified containment protocols and expanded reporting. The Los Angeles Times says the 85-page plan covers loss-of-control disclosures. Proposed caps address recursive self-improvement: AI designing successors without human input.
  • Anadolu says the California Office of Emergency Services would investigate after shutdowns and update safety guidance.

Sources

Sources