California Bill Would Require Disclosure of AI Safety Computing

A California legislative proposal would require artificial intelligence developers to disclose the computing resources used for safety testing of their models. The bill marks a formal step toward creating auditable records that connect safety investment to the scale of model development and the risks posed by deployment.

The measure signals growing demand among policymakers and industry observers for verifiable evidence that safety protocols expand in proportion to a system’s capabilities and potential harms. By mandating compute disclosures tied specifically to safety evaluations, the proposal aims to make testing practices more transparent to regulators, insurers, and other stakeholders who assess liability and operational risk.

The requirement reflects a broader push to establish documentation standards as AI systems become more powerful and widely integrated across sectors including education, real estate, construction, science, and communications. If enacted, the rule would give government agencies and legal professionals a concrete metric for evaluating whether developers are dedicating sufficient resources to risk mitigation before releasing advanced models.