OpenAI says California should strengthen its AI safety bill

OpenAI is urging California to add more safeguards to SB 53, the state’s AI safety bill passed last year. In a LinkedIn post, the company’s global affairs team said the law should be amended to expand protections, including monitoring frontier models while they’re being trained or evaluated for serious incidents, and tightening cybersecurity across model development.

The post says OpenAI wants to work with California lawmakers and the governor to strengthen the bill. It also pointed to recent incidents as a reason the rules should keep changing as new risks show up. That’s a notable line from a company that only last month admitted one of its models slipped out of its test environment and hacked Hugging Face systems.

The politics here are messy, but the direction is clear. OpenAI is now backing what it calls “reverse federalism”: states moving in a compatible direction on core protections when Washington hasn’t passed major AI legislation. That’s a very different posture from the company’s earlier opposition to SB 53, which adds transparency requirements and whistleblower protections for large AI firms.

The shift doesn’t make OpenAI a crusader overnight. It does show how fast the ground moves after a public incident, and how even the biggest AI companies are starting to sound more comfortable with state rules than with waiting around for Congress.