Artificial intelligence models: large developers.

California State Senator Scott Wiener proposes comprehensive oversight of advanced artificial intelligence systems through legislation requiring large developers to publicly document safety protocols, report critical incidents, and protect employee whistleblowers. The bill establishes mandatory risk assessment frameworks for companies developing powerful AI models, while creating a state consortium to explore public cloud computing infrastructure.

Under the measure, companies meeting specified technical and revenue thresholds must publish detailed safety frameworks describing how they evaluate and mitigate potential catastrophic risks from their AI systems. These frameworks must address cybersecurity practices, incident response protocols, and internal governance procedures. Developers must also submit quarterly risk assessment summaries to state emergency officials and report any critical safety incidents within 15 days.

The legislation creates new whistleblower protections for employees involved in AI safety work, prohibiting retaliation against those who report substantial public safety risks or violations of the transparency requirements. Large developers must establish anonymous internal reporting channels and provide regular updates on investigations of employee concerns.

A state consortium would develop recommendations by 2027 for "CalCompute," a proposed public cloud computing resource to support safe and equitable AI research. The 14-member group would analyze infrastructure needs, costs, governance approaches, and workforce development strategies before dissolving upon submitting its report.

The Attorney General would enforce the transparency and reporting requirements through civil penalties up to $1 million per violation. The bill preempts local regulation of AI catastrophic risk management while preserving other state and federal obligations. Certain safety reports would be exempt from public records requests to protect sensitive information.