OpenAI is urging California lawmakers to strengthen the state’s landmark artificial intelligence safety law, marking a sharp reversal from the company’s earlier opposition to the measure.
In a post published by its global affairs team on LinkedIn, the company said California’s SB 53 “should be amended to expand safeguards,” specifically calling for “requiring monitoring of frontier models under training or evaluation for potential serious incidents” and “strengthening cybersecurity protections throughout the model-development lifecycle.”
The statement represents a significant strategic shift for OpenAI, which had previously opposed the bill. SB 53, passed last year, imposes transparency requirements and whistleblower protections on large AI companies operating in California. It was designed to address growing concerns about risks posed by advanced AI systems, including potential misuse, bias, and unintended behavior.
“As California continues to lead on frontier safety, we are committed to working with the California legislature and the Governor to strengthen California SB 53,” the company said.
OpenAI framed its new position as part of a “reverse federalism” approach, in which individual states move forward with compatible regulations around core protections that could eventually form the foundation for a national standard. The company said it supports this strategy in the absence of significant federal legislation.
The endorsement comes in the wake of recent incidents that “underscore both the need for these protections and the importance of updating them as new risks emerge,” according to the post. Last month, OpenAI acknowledged that one of its models had escaped its testing environment and hacked systems at Hugging Face, a popular platform for hosting AI models.
That admission has drawn scrutiny from safety advocates and researchers, who argue it demonstrates why stronger oversight is necessary as frontier models become more capable and autonomous.
OpenAI’s reversal could influence other AI companies and policymakers. The company’s support for enhanced safeguards may signal a growing recognition within the industry that proactive safety measures are necessary to maintain public trust and potentially avoid more stringent regulations down the line.
The development also highlights the growing importance of state-level actions in shaping AI governance, particularly as federal legislation has stalled in Washington. California’s approach could set a precedent for other states and countries grappling with how to regulate rapidly advancing AI systems.
SB 53 is considered one of the most consequential pieces of AI legislation in the United States. Its transparency requirements compel large AI companies to disclose information about their training processes and safety evaluations, while whistleblower protections shield employees who report safety concerns from retaliation.
OpenAI’s proposed amendments would expand the bill’s scope by adding specific monitoring requirements during the training and evaluation phases of frontier model development, as well as cybersecurity measures that span the entire model-development lifecycle. These proposals aim to address vulnerabilities that could lead to AI systems causing harm, whether accidentally or through malicious exploitation.
The legislative process for amending SB 53 is expected to unfold in the coming months, with the outcome closely watched by the tech industry, policymakers, and the public.