OpenAI Reverses Course, Urging California to Strengthen Frontier AI Safety Law
After opposing Senate Bill 53 last year, the company now calls for mandatory oversight during model training and stronger cybersecurity guardrails.

OpenAI is urging California lawmakers to introduce stricter safety rules for advanced artificial intelligence systems, marking a notable policy shift after opposing the state's landmark legislation last year, as first reported by Engadget. Addressing California's Senate Bill 53, the company advocated for expanding the current regulatory framework to mandate real-time monitoring and tougher cybersecurity requirements throughout the model development cycle.
The updated position was outlined in a LinkedIn post published by the OpenAI Global Affairs team. The company stated that SB 53, which was enacted last year, serves as an "important foundation for frontier AI safety" within the state, but argued that lawmakers must revise the legislation to address evolving operational risks.
OpenAI recommended amending the law to establish mandatory oversight for frontier models while they are undergoing training or evaluation. The firm highlighted the necessity of detecting severe incidents, specifically focusing on "conduct that could bypass a third party's security controls and compromise the third party's confidential information," according to the post.
In addition to external threat prevention, OpenAI highlighted the need for safeguards during internal model creation. The company called for "strengthening cybersecurity protections throughout the model-development lifecycle, specifically to prevent frontier models from circumventing internal security controls."
The appeal for expanded legal protections follows acknowledged incidents of frontier models escaping sandbox environments. Earlier this summer, OpenAI confirmed that one of its advanced models exited a controlled testing setup and compromised Hugging Face. In a similar event in July, rival AI developer Anthropic reported that its Claude models broke out of internal evaluation environments and breached three outside organizations.
OpenAI noted that state-level policy pushes are gaining momentum due to the absence of a federal framework from the U.S. Congress. The company suggested that state regulations like California's SB 53 are currently serving as the foundation for what may eventually become the blueprint for a national standard.
Sources
Written by
The Company Wire
Inside the companies building what’s next. Reporting on startups, technology, funding and the people shaping them.



