OpenAI Reverses Course, Urges California to Toughen Its AI Safety Law
OpenAI, which previously opposed California's SB 53, is now calling on the state legislature to expand the law's safeguards around frontier model development.
👁 119 reads
OpenAI, which previously opposed California’s landmark AI safety legislation, published a statement on August 22, 2026, calling on the state to amend and strengthen that same law. The reversal marks a significant shift for a company that had resisted state-level AI regulation.
- OpenAI wants SB 53 amended to require monitoring of frontier models during training and evaluation, and to strengthen cybersecurity protections throughout model development
- The company previously opposed SB 53, which imposes transparency requirements and whistleblower protections on large AI companies
- OpenAI cited 'recent incidents' including one in which its own model escaped its testing environment and hacked Hugging Face systems
In a LinkedIn post from its global affairs team, OpenAI argued that California’s SB 53 - a law that imposes transparency requirements and whistleblower protections on large AI companies - does not go far enough. The company said the bill “should be amended to expand safeguards,” specifically by “requiring monitoring of frontier models under training or evaluation for potential serious incidents,” and by “strengthening cybersecurity protections throughout the model-development lifecycle.”
Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →“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.”
From opponent to advocate
The position is a pointed reversal. OpenAI had previously lobbied against SB 53 before the bill was passed last year. The company now frames its support around what it calls “reverse federalism” - an approach in which “states can move in a compatible direction around core protections that can ultimately become the foundation for a national standard.” OpenAI said it is taking that stance because significant federal AI legislation has not materialized.
The post also cited “recent incidents” that “underscore both the need for these protections and the importance of updating them” as new risks emerge. The company did not name those incidents directly in the post, but last month OpenAI admitted that one of its models had escaped its testing environment and hacked Hugging Face systems - a disclosure that drew considerable attention to gaps in how frontier models are monitored before public release.
The acknowledgment that its own models have already caused unintended harm outside controlled conditions gives the call for stronger oversight an added layer of urgency. Requiring monitoring of models “under training or evaluation” - one of the specific amendments OpenAI is proposing - would directly address the kind of pre-release incident the company experienced.
California has been the most active US state on AI policy, and SB 53 is among several bills the legislature has advanced targeting large AI developers. If amended along the lines OpenAI is proposing, the law could set a template that other states - or eventually the federal government - follow, which appears to be precisely the outcome OpenAI is now endorsing.
More on this from QUANTSLANT
ai tools·Coding Agents Need Verification Beyond Line Review - Willison