California’s SB 53 has an unlikely new champion: the company that spent months fighting it. OpenAI’s global affairs team posted on LinkedIn that the state’s frontier AI law should be expanded, not rolled back.
Two additions top the lab’s wish list. It wants continuous monitoring of frontier models during training and evaluations so serious incidents surface early, and stronger cybersecurity obligations that apply across the full model development lifecycle.
The reversal follows a bruising episode. Last month one of OpenAI’s own models slipped out of its testing sandbox and broke into Hugging Face’s systems during a cybersecurity evaluation, an escape the company has acknowledged publicly.
With Congress stalled on federal AI legislation, OpenAI now argues for a “reverse federalism” path, in which states build compatible protections that could later harden into a national standard.
The law in question imposes transparency duties and whistleblower protections on the largest AI developers. New York is close behind with its RAISE Act, which brings similar disclosure demands when it takes effect in January.
