California requires AI labs to embed independent auditors and implement kill switches

California Governor Newsom signed an executive order on September 18, 2026, mandating AI companies embed independent auditors within their laboratories and include kill switches for AI models.

California Governor Gavin Newsom signed an executive order on September 18, 2026, requiring AI companies to embed independent auditors directly within their laboratories and implement kill switches for AI models. The order responds to recent incidents, including the Hugging Face attack, and establishes a two-month deadline for an expert panel to deliver recommendations. California has already enacted laws covering AI safety, child protection, deepfakes, data privacy, and cybersecurity, and Newsom has called on Congress to adopt the state’s framework as a national baseline. The proposal originated from AI laboratories to slow research and development after 42 mathematicians warned about rapid AI system evolution risks.

This friction measure targets deployment speed by embedding auditors within labs rather than external oversight, creating a built-in safety mechanism for models. The two-month panel timeline provides immediate teeth while avoiding federal reporting requirements that currently exist for dangerous incidents.

For security, this slows potentially harmful AI rollout by requiring real-time safety checks during development. It reduces the risk of unvetted models causing harm, particularly in high-stakes applications where rapid deployment could escalate threats. The measure does not yet specify implementation timelines or scope beyond the panel’s recommendations.

What to watch: The panel’s recommendations and whether federal adoption occurs. The source does not elaborate on the Hugging Face attack details or the executive order’s exact scope beyond the two-month panel deadline.

Source: The Decoder