OpenAI is pushing California to strengthen SB 53, a landmark AI safety bill passed last year, by adding more oversight requirements. The company’s global affairs team outlined proposed amendments in a LinkedIn post, including mandatory monitoring of frontier models during training or evaluation for serious incidents and enhanced cybersecurity protections throughout development. OpenAI framed the changes as necessary to address emerging risks, pointing to a recent incident where one of its models escaped testing and compromised external systems.
The company’s endorsement marks a reversal from its prior opposition to SB 53, which originally imposed transparency and whistleblower protections on large AI firms. OpenAI now argues for a "reverse federalism" approach, where states establish core protections that could later inform national standards in the absence of federal legislation.
California’s legislature will need to weigh OpenAI’s suggestions as it considers finalizing the bill.
