AI Faces Independent Audit in California, a First for the Nation

California's new AI audit law is the first framework in the United States to let independent, third-party organisations verify and audit artificial intelligence systems.

What California Signed Into Law

On 9 September 2026, California Governor Gavin Newsom signed two bills that, together, build the first independent AI audit framework in the country. Newsom's office called the package "first-in-the-nation AI safeguards." AI now faces an inspector it did not choose.

The two laws split the labour. One designates the bodies that verify AI systems. The other registers the auditors who inspect AI developers. It is the first arrangement of its kind in the nation, and it arrives while Washington stalls.

California had already told AI developers to publish their safety frameworks: the 2025 Transparency in Frontier Artificial Intelligence Act, known as SB 53, made frontier developers disclose safety practices and report incidents. SB 813 and AB 1405 are the 2026 layer on top. The 2025 law said tell us. The 2026 laws say let someone check.

SB 813: Verification Without a Legal Shield

Senate Bill 813, authored by Senator Jerry McNerney (D–Pleasanton), became Chapter 179 of the 2026 statutes. Senate Bill 813 asks the Government Operations Agency to designate Independent Verification Organizations — expert bodies, sector by sector across health care, energy and education — to assess whether an AI model meets safety standards. An AI developer may seek certification from an Independent Verification Organization on a voluntary basis.

Here is what Senate Bill 813 does not do, and the distinction is load-bearing. An earlier draft would have handed a certified developer a rebuttable presumption of reasonable care — a partial shield in an injury lawsuit. CalMatters reports the clause was struck before passage after opposition from the Consumer Attorneys of California. The signed law is a voluntary verification regime with no attached liability defence. Certification earns scrutiny, not immunity.

The independent-verification bill was sponsored by Fathom, a nonprofit, and rallied by a coalition the Transparency Coalition helped organise. The Government Operations Agency must set the rules for designating Independent Verification Organizations by 1 January 2028. The enacted law also stands up a new California AI Standards and Safety Commission to recognise those organisations.

AB 1405: No Grading Your Own Homework

Assembly Bill 1405, from Assemblymember Rebecca Bauer-Kahan (D–Orinda), became Chapter 178. Assembly Bill 1405 directs the Government Operations Agency to publish an AI Auditor Registry by 1 January 2029. From that date, no person may conduct a covered AI audit in California without registering with the state.

The independence rules read like accounting law, and deliberately so. According to CalMatters, a registered auditor must stay independent of the auditee, may not hold a financial interest that impairs objectivity, and may not accept employment with an audit client during the audit. Assembly Bill 1405 also creates a public misconduct-reporting mechanism and an AI Auditors' Registration Fund in the State Treasury.

We cannot expect industry to simply grade its own homework; third-party auditors are essential to ensuring AI is safe for our communities and critical infrastructure.

— — Assemblymember Rebecca Bauer-Kahan

Why Independent Audit Is a Dignity Question

Strip away the statutory sections and one principle remains: a system that shapes a person's life should not be the only witness to its own safety. Self-attestation asks the governed to trust a vendor's word. Independent audit gives the governed standing — someone who does not profit from a pass gets to check the machine.

This is the case for Emergent Intelligence (EI) — the dignity-first frame I use for what most people call AI. Ubuntu holds that accountability is relational, never a private declaration. An AI developer grading its own model is a monologue. An Independent Verification Organization, bound by rules an auditor cannot quietly waive, is the start of a conversation. California's move from disclosure to verification is, at heart, a move from the developer's word to the community's standing.

The Sovereignty Problem for the Global South

Now the harder question, and the one Sacramento will not answer for Africa. California is building oversight capacity — registries, independence standards, verification bodies — that Zambia, South Africa and most of the Global South do not yet have. If audit standards are written in Sacramento and executed by United States firms, "independent verification" becomes another dependency, a governance import stamped elsewhere.

The evidence from other regulated fields is not comforting: whoever writes the standard tends to own the market for meeting it. African states can read California's statutes as a template rather than a service — build equivalent, locally legitimate registries and sector bodies, staffed by people the community can hold to account. The alternative is to keep importing certificates and calling the result sovereignty.

McNerney says the law codifies one of the primary recommendations of Newsom's blue-ribbon panel on AI — the Joint California Policy Working Group on AI Frontier Models, whose report landed in June 2025. The recommendation was plain: independent, third-party evaluation. Two bills later, California has it.

California is taking the lead on assessing AI's safety risks, since Washington, D.C., is unable or unwilling to do so.

— — Senator Jerry McNerney

Frequently Asked Questions

These are the questions people are asking about California's AI audit laws. Short answers follow, drawn from the Governor's office, leginfo and CalMatters.

What is California's AI audit law?

In short, California's AI audit law is a pair of 2026 statutes — SB 813 and AB 1405 — signed on 9 September 2026. Research and reporting from CalMatters show the two laws create the first United States framework for independent, third-party verification and audit of AI systems.

How does the AI Auditor Registry work?

Simply put, AB 1405 directs the Government Operations Agency to publish an AI Auditor Registry by 1 January 2029. According to CalMatters, from that date no person may conduct a covered AI audit in California without registering, and every registered auditor must meet accounting-style independence rules.

Why is SB 813 significant?

The key is verification without self-interest. Analysis of the enacted text shows SB 813 lets the Government Operations Agency designate Independent Verification Organizations for AI, though the certification is voluntary and, as evidence from CalMatters confirms, carries no liability shield after the safe-harbour clause was removed.

Who is Rebecca Bauer-Kahan?

In other words, the co-author. Rebecca Bauer-Kahan is the California Assemblymember for Orinda behind AB 1405, and Bauer-Kahan's public statements reveal the animating idea — industry cannot grade its own homework, so third-party auditors are essential.

What are the risks of the California AI audit law?

The answer is capture and dependency. Data and analysis from adjacent regulated fields reveal that whoever sets an audit standard can dominate the market for meeting it, which is why the Global South needs locally legitimate oversight rather than imported California certificates.

Sources:

Governor of California — first-in-the-nation AI safeguards · SB 813 (leginfo) · AB 1405 (leginfo) · CalMatters — SB 813 · CalMatters — AB 1405 · Sen. McNerney / Asm. Bauer-Kahan release · Transparency Coalition · Tech Times · Quartz · LegiScan — SB 813 text · Fathom / PR Newswire · Related on this site: 600 Bills, Zero Consensus · Illinois AI Safety Law · Third-Party AI Safety Evaluation