Signal

California creates an independent-verification framework and registry for AI auditors

California Governor Gavin Newsom signed SB 813 and AB 1405 on September 9, creating a framework for independent verification organizations and a state registry for AI auditors. The measures establish governance infrastructure for third-party AI assessments, while practical impact will depend on implementation.

2 min read
Official California government graphic announcing new AI safeguards
California's official announcement graphic for the new AI-safeguards legislation · Credit: Office of Governor Gavin Newsom View source

California Governor Gavin Newsom signed two linked AI-oversight bills on September 9. Senate Bill 813 establishes a framework for independent verification organizations, while Assembly Bill 1405 creates a registry for AI auditors and standards intended to protect their independence and integrity.

A new layer between developers and regulators

The central change is institutional. Instead of relying only on developers' own safety claims or direct government inspections, California is creating a route for qualified third parties to assess AI systems and models against state requirements. The governor's announcement describes the two laws as complementary parts of a framework for independent evaluation.

SB 813 covers independent verification organizations and the standards expected of them. AB 1405 focuses on the auditor registry, including independence, transparency, and integrity requirements. State Senator Jerry McNerney's legislative summary says the verification framework is designed to require relevant expertise, credible methods, and separation from the entities being assessed.

Why the distinction matters

An auditor registry does not by itself establish that a model is reliable, unbiased, or secure. Its value depends on who qualifies, how conflicts of interest are handled, which tests become accepted, and whether reports give regulators and the public enough evidence to evaluate compliance. Those implementation details will determine whether the framework produces rigorous scrutiny or merely another certification layer.

The measures build on California's broader AI regime, including the state's 2025 frontier-model transparency law. They also arrive as governments debate how to verify increasingly consequential claims about model safety, security, and behavior without requiring every regulator to operate a frontier-model testing laboratory.

What can be verified now

The governor's office confirms that both bills were signed and describes their assigned roles. Legislative materials from Senator McNerney independently document the intended third-party assessment structure and the bipartisan votes that sent SB 813 to the governor. What cannot yet be measured is the framework's enforcement quality, market adoption, or effect on developer behavior because those depend on implementation after enactment.

Sources