California Enacted Audit Rules for AI Developers
The state's new regulatory framework mandates independent verification, while private firms move to scale internal safety testing.
Updated on Oct. 11, 2026 in Artificial Intelligence
Live Poll
Should the federal government mandate independent audits for advanced artificial intelligence models?
On Sept. 9, 2026, California enacted Senate Bill 813 and Assembly Bill 1405, establishing a formal state registry for AI auditors and verification organizations. These laws aim to create external oversight for model development, contrasting with current industry-led safety initiatives.
Why it matters
The legislation formalizes oversight to address the current lack of industry-wide standards for AI risk management and reporting. This move shifts the focus toward mandatory accountability as federal efforts, such as the FRONTIER Act, continue to take shape.
METR reported $13.64 million in gross receipts for the 2024 tax year and secured $71 million in funding commitments through August 2026. These figures serve as benchmarks for the scale of current non-profit evaluation efforts compared to the $1 billion investments recently announced by private industry players.
The players
Anthropic
An AI research and deployment company focused on building large language models with a heavy emphasis on safety and constitutional engineering.
Accenture
A multinational professional services company providing consulting and technology strategy, now increasingly focused on AI implementation and risk management.
Gavin Newsom
The Governor of California who signed the state's latest legislative package into law on Sept. 9, 2026.
METR
A research organization specializing in evaluating the safety and performance of artificial intelligence systems.
The details
The new laws require developers to engage with independent entities that perform rigorous stress testing on model architectures. Anthropic plans to facilitate this by embedding evaluators from Accenture directly into its workforce, providing them with internal system access to identify unauthorized behavior. These evaluations are intended to supplement model cards—standardized documents detailing a model’s training data and performance limitations—as requested in federal legislative proposals.
Timeline
May 1 to Dec. 31, 2024: METR's fiscal 2024 tax year.
July 23, 2026: The FRONTIER Act (H.R. 9925) was introduced in the House.
Aug. 14, 2026: METR concluded its six-month funding period.
Sept. 9, 2026: California Governor Gavin Newsom signed the new AI audit legislation.
Sept. 18, 2026: Anthropic announced its partnership with Accenture.
The Tech Race
California's new legislative framework aligns with the requirements proposed by the FRONTIER Act (H.R. 9925). These state laws effectively establish a pilot for the federal risk management and audit requirements that remain under debate in the House.
Industry participants in California will need to register with the state and facilitate third-party audits of their model architectures. These measures aim to standardize transparency, though the practical impact on software workflows and development timelines remains to be seen as the registry launches.
The takeaway
California's new audit requirements signal a pivot toward mandatory external validation for AI developers. Stakeholders should track the establishment of the state auditor registry, which will dictate how firms like Anthropic and their partners must document system safety for the foreseeable future.
Further reading
Explore the broader regulatory landscape in our Artificial Intelligence section.
Source note: This article includes information reported by TokenPost.
Live Poll
Should the federal government mandate independent audits for advanced artificial intelligence models?
