OpenAI is publicly backing changes to California's SB 53, saying the state's frontier AI safety framework should be expanded rather than held fixed.

In a post from its global affairs team, the company said states can help form a national baseline while Congress continues debating federal AI legislation. OpenAI described California's law as an "important foundation" because it covers risk assessment, transparency, incident reporting and security for developers of the most capable models.

The company is now asking lawmakers and the governor to strengthen the framework. Its suggested additions include monitoring frontier models during training or evaluation for serious incidents, with a focus on behavior that could bypass third-party security controls or compromise confidential information. OpenAI also said cybersecurity protections should be strengthened throughout the model-development lifecycle.

The position is notable because OpenAI had previously opposed SB 53, according to TechCrunch. The legislative text identifies the measure as SB 53, "Artificial intelligence models: large developers," and describes requirements for large frontier developers to publish a frontier AI framework and submit certain safety-related information to state emergency officials.

For AI infrastructure companies, the practical signal is that policy pressure is shifting from generic transparency toward operational controls around frontier-model development. OpenAI's statement does not settle how those controls should be implemented, but it gives California legislators explicit support from one of the industry's largest labs for adding more monitoring and security obligations.