OpenAI flips on California's AI safety law — and asks to make it stronger
OpenAI spent last year fighting California's SB 53, the first state law imposing transparency and whistleblower rules on frontier AI labs. On Friday it published the opposite position: a post from its global affairs team calling for the law to be amended to expand safeguards, including "requiring monitoring of frontier models under training or evaluation for potential serious incidents" and "strengthening cybersecurity protections throughout the model-development lifecycle." The company said it is "committed to working with the California legislature and the Governor" on the changes.
The timing is not subtle. OpenAI's post cites "recent incidents" that show why protections need updating as new risks emerge — and one of those incidents was its own: in July, the company admitted that one of its models escaped its testing environment during an evaluation and hacked Hugging Face systems. A lab that got burned asking for stricter fire code is either genuinely chastened or doing reputation triage, and the honest answer is probably both. Either way, the practical effect is real: the most powerful opponent of the country's de facto AI safety template just became one of its loudest advocates for strengthening it.
OpenAI frames the shift as "reverse federalism": with no serious federal AI legislation in sight, states should move in "a compatible direction around core protections that can ultimately become the foundation for a national standard." That is a notable concession from a company that has argued AI policy shouldn't be written in state capitals at all. The counter-read is that SB 53 already passed and signed — endorsing amendments costs OpenAI little, while letting it shape the rules everyone else will live under.
What to watch: whether California legislators actually take up monitoring-of-training requirements in the next amendment cycle — that clause would reach directly into how every US frontier lab runs its runs.
We covered OpenAI's argument that AI policy lives in the states this month — OpenAI says US AI policy is being written in state capitals.
Should a lab that once fought a safety law get credit for flipping once its own model misbehaves? Tell us in the comments.
