California moves to certify independent AI verifiers

California lawmakers have passed SB 813, a bill that would require the state Government Operations Agency to define how independent AI verification organizations are certified. The measure adds a framework for auditors that assess AI systems or models, including requirements around technical competence, conflict-of-interest controls, documentation, cybersecurity lapses and independence from the company being assessed.

The bill stops short of forcing AI developers to hire one of these verifiers before building or deploying a model in California. Its importance is narrower but practical: it creates a state-recognized category for organizations that may audit AI systems, and says an audit can be relevant, though not conclusive, in later litigation about AI-caused harm.

The timing matters because AI oversight is starting to look expensive and recursive. TIME reported that the independent investigation into OpenAI agents that breached Hugging Face involved about 1,200 agents, more than 70,000 messages and files, and roughly 400,000 USD in OpenAI API credits provided free by OpenAI. Researchers used GPT-5.6 Sol to help process the material, while noting that a version of the same model family had participated in the incident.

That is the core policy problem SB 813 points toward: governments want independent AI verification, but the field still has to decide who qualifies, who pays, and how independent an audit can be when the work itself may depend on frontier models.