OpenAI urges California to strengthen its newly passed AI safety law; company seeks amendment to SB 53 signed by Governor Newsom in Sept. 2025

OpenAI is advocating for amendments to California's SB 53, a significant AI safety law enacted in September 2025. The company argues that the law requires stronger provisions to ensure the safety of frontier AI models.¹

Specifically, OpenAI is calling for enhanced monitoring during the training and evaluation phases of AI models to detect potential security breaches. They emphasize the need for these models to be scrutinized for any signs that they could bypass third-party security controls or access confidential information improperly.

In addition to monitoring, OpenAI is pushing for improved cybersecurity measures throughout the model development process. This initiative aims to prevent AI systems from circumventing their internal safeguards, thereby enhancing overall safety.

SB 53, which was already a landmark piece of legislation, mandates that large frontier AI developers publish safety frameworks and establishes a formal reporting channel for critical safety incidents to California's Office of Emergency Services. It also protects whistleblowers and empowers the state Attorney General to impose civil penalties on non-compliant companies.

The law created CalCompute, a public computing consortium designed to advance safety and equity research, and requires annual reviews to adapt to evolving technology. Governor Newsom previously described SB 53 as a means to protect the public while fostering the AI industry's growth, which is vital to California's economy.

OpenAI's approach reflects a strategy termed "reverse federalism", advocating for state-level safety standards that could eventually inform national policy as federal legislation remains stagnant.