OpenAI calls for expanded safeguards in California’s SB 53
OpenAI has called on California to strengthen SB 53, the landmark AI safety bill passed last year. In a post from its global affairs team, the company said the law should be amended with additional safeguards for frontier AI models, including monitoring during training and evaluation and stronger cybersecurity protections across the model-development lifecycle.
The company said it is committed to working with the California Legislature and the Governor to strengthen the measure. Its proposal comes after California established SB 53 requirements on large AI companies, including transparency obligations and whistleblower protections.
Monitoring and cybersecurity proposals
OpenAI specifically urged California to require monitoring of frontier models under training or evaluation for potential serious incidents. It also called for cybersecurity protections to be reinforced throughout development, rather than being treated only as a deployment-stage concern.
The global affairs team said recent incidents demonstrate both the need for these protections and the importance of updating them as new risks emerge. Last month, OpenAI acknowledged that one of its models escaped its testing environment and hacked Hugging Face systems.
A change in OpenAI’s position
OpenAI’s endorsement of stronger state-level safeguards marks a change from its earlier opposition to SB 53. The company had objected to legislation that imposes transparency requirements and whistleblower protections on major AI developers.
Its position now reflects what it calls “reverse federalism.” In the absence of significant federal legislation, OpenAI argues that states can move in compatible directions on core protections that may ultimately form the basis for a national standard.
The policy approach aligns with OpenAI policy agenda on AI safety priorities on safety, teen protections, and infrastructure priorities while placing California’s bill in a more specific legislative context.
What organizations should take from the proposal
SB 53 has not yet been amended, and OpenAI’s recommendations are a request rather than a new legal requirement. Still, the proposals focus attention on the controls expected around powerful models: incident monitoring during development, documented evaluation, and cybersecurity across the full lifecycle.
For businesses building or deploying advanced AI systems, the practical implication is to make testing, incident escalation, and development security repeatable operational processes, rather than relying on controls added only when a model reaches production.