Anthropic pushes state-by-state ratchet on AI safety rules

Two frontier labs are running opposite plays on the same regulatory board. Politico reports that Anthropic is deliberately backing progressively tougher state AI safety bills, while OpenAI global-affairs chief Chris Lehane pursues what he openly calls 'reverse federalism,' getting states to replicate each other's laws so a workable standard hardens into a de facto national one before Congress moves.

Anthropic's move list is now on the record. It endorsed California's SB 53, signed in late September 2025 as the first US statute focused squarely on AI safety, then New York's RAISE Act, which Governor Hochul signed on December 19. It backed an Illinois bill requiring independent, third-party audits of large frontier AI developers' safety practices. Now the company is backing a Massachusetts measure that would apply to labs with more than $500 million in annual AI-derived revenue or more than $1 billion in AI R&D spend, require independent evaluators to review catastrophic-risk mitigations at least every six months, and let the state attorney general seek court orders to enforce it. Anthropic's head of US state and local government relations, Cesar Fernandez, calls it 'the strongest AI safeguards in the country.'

Why it matters if you don't lobby for a living: the two labs are effectively racing to set the ceiling of what state AI law can demand. Anthropic's public line is that these rules formalize practices it already follows, which is a comfortable position for a lab that has already built the compliance muscle and a much less comfortable one for smaller frontier players who suddenly face six-monthly external audits and a half-billion-dollar revenue trigger. OpenAI's counter-play, per Lehane, is to keep the mirror-state template closer to what large incumbents can live with, then let those blue-state rules substitute for a federal one.

Take the specifics as reported, not settled. The retrieved coverage doesn't tell you which Massachusetts language will actually survive final passage inside the broader economic development bond bill it was attached to, how state AGs will staff enforcement, or whether the mirror-and-ratchet dynamic holds once a red state weighs in. What it does tell you is that the frontier-lab lobbying game has moved out of DC and into statehouses, and the labs writing the ratchet are the ones best placed to live under it.