California’s new AI laws, signed September 9–10, 2026, do not ban a single model, cap a single parameter, or impose liability on a single developer. They build the regulatory plumbing that future, harder rules will run through.
Governor Gavin Newsom signed two bills into law. SB 813, authored by Sen. Jerry McNerney (D-Pleasanton), creates the California Artificial Intelligence Standards and Safety Commission and a certification pathway for independent verification organizations (IVOs). AB 1405, authored by Asm. Rebecca Bauer-Kahan (D-Orinda), establishes a state AI auditor registry. Together, they define who gets to audit AI systems in California and what those auditors must do to qualify.
What the Laws Actually Do
The two bills create a shared infrastructure for AI oversight, but neither one mandates that any AI system be audited today.
SB 813 tasks the Commission with developing voluntary AI safety standards across two tiers: maximum compliance and advanced safety. It also certifies IVOs that can test AI systems for safety, reliability, and security. The bill is structurally contingent on AB 1405 passing; neither framework functions without the other.
AB 1405 creates the registry within the Government Operations Agency. Auditors must meet independence standards, observe cooling-off periods before taking employment with companies they have audited, and document their findings transparently.
Bauer-Kahan was precise about the bill’s scope at a CalMatters hearing. AB 1405 “does not require auditing of AI systems and models” but “simply creates accountability and transparency requirements for AI auditors that operate in California.”
The framework is voluntary for now. When California eventually enacts sector-specific rules covering AI in hiring, credit scoring, or health services, this infrastructure becomes the automatic backbone those rules plug into.
Why the Industry Said Yes
Anthropic and OpenAI endorsed the bills before Newsom signed them, and their reasons reveal exactly what this framework is and is not.
Chris Lehane, OpenAI‘s Chief Global Affairs Officer, framed the moment in a statement published ahead of the signing: “We’ve reached a new chapter in AI capabilities, and that demands a new chapter for AI policy.”
Artificial intelligence holds extraordinary promise, but it must be developed and deployed with meaningful safeguards to protect the public.
Governor Gavin Newsom, official signing statement
What frontier labs get from this framework is an official, certified channel to demonstrate safety compliance. That is useful for regulated sectors and public trust, without any obligation to slow model development. Voluntary standards plus certified third-party auditors is precisely what companies like Anthropic and OpenAI have been advocating at federal and international levels. California just gave that preference a legal address.
What It Means for You
No AI system your employer, insurer, or bank uses is guaranteed to have been audited under this framework yet.
IVO certification is expected by January 1, 2028. The auditor registry becomes fully operational, with registration mandatory, by January 1, 2029. The near-term effect is institutional: commissions formed, criteria written, registries built.
Former Anthropic researcher Jacob Coxon has publicly warned that AI leaders fear they may be hastening catastrophe. The statutes themselves, however, address process and certification, not capability limits. Newsom declined to endorse apocalyptic framing when asked about those concerns. He called on the country to enact “robust, national regulations that match the urgency of this moment,” according to Politico.
How meaningful this framework becomes depends almost entirely on the Commission’s rigor in writing standards. The difference between a genuine safety regime and a rubber-stamp label companies collect with minimal friction will be written in that fine print.
