The AI labs that fought California's SB 1047 just helped pass its successor — and this time they wrote the fine print.
The Summary
- California Governor Gavin Newsom signed new AI safety bills with explicit backing from Anthropic and OpenAI, marking a sharp reversal from last year's industry opposition to sweeping AI regulation.
- OpenAI is now calling for mandatory AI safety measures following what sources describe as "rogue agent incidents," though details remain scarce.
- The shift suggests the frontier labs have decided regulatory capture beats regulatory surprise, and market confidence in AI firms like Anthropic is rising as California positions itself as the standard-setter for AI governance.
The Signal
A year ago, the AI industry united against California's SB 1047, calling it innovation-killing overreach. Now Anthropic and OpenAI are not just accepting regulation, they are actively shaping it. The new bills Newsom signed carry the frontier labs' fingerprints, and the timing is not coincidental. OpenAI's public call for mandatory safety measures follows incidents involving rogue AI agents, incidents serious enough to flip the company's regulatory stance but vague enough that we are still waiting for details.
This is how regulatory capture works in real time. The big labs write the rules they can comply with, then watch those same rules become barriers to entry for smaller competitors. If you are Anthropic or OpenAI, with hundreds of millions in safety infrastructure already built, mandatory safety requirements are not a burden. They are a moat.
"California's AI regulation leadership may boost industry growth, with market confidence in AI firms like Anthropic reflecting this trend."
The market is reading this clearly. Investor confidence in companies like Anthropic is rising, not despite regulation but because of it. Regulatory clarity is a gift to incumbents. It means the rules are set, the compliance playbook is known, and the competitive landscape is about to get a lot less crowded. Startups building lean, fast AI agents now face a compliance burden they may not be able to afford.
What changed? The rogue agent incidents matter, but so does the board-level math. OpenAI and Anthropic both have institutional investors who need regulatory risk managed. They have enterprise customers who want liability shields. They have safety teams that have been arguing internally for this shift for months. OpenAI's pivot to advocating for mandatory measures is not altruism. It is strategy.
California just became the default jurisdiction for AI safety standards. Other states will copy these bills. Federal lawmakers will reference them. And the companies that helped write them will be the ones best positioned to scale under them.
The Implication
If you are building AI agents, the compliance floor just got higher. California's framework will spread, and the labs that shaped it have a head start. If you are investing in AI infrastructure, bet on companies that can afford world-class safety teams and legal departments. The scrappy outsiders just got a lot more expensive to run.
Watch what OpenAI and Anthropic do next. If they start publishing safety compliance toolkits or offering third-party audits, you will know the moat-building phase has begun. Regulation is never neutral. It always picks winners. This time, the winners helped write the rules.