Inside Anthropic's state-by-state plan to ratchet up AI rules

Anthropic is pursuing a state-by-state strategy to push for increasingly stricter AI safety regulations, contrasting with rival OpenAI's push for a uniform national framework mirrored across states, the company's head of U.S. state and local government relations, Cesar Fernandez, said in an interview with POLITICO on Tuesday. Fernandez said Anthropic is not looking to support the same bill in every state but seeks legislation that meaningfully raises the bar on safety for the most capable AI systems. The approach differs from OpenAI's "reverse federalism" strategy, described by its top lobbyist Chris Lehane, which aims to build a national framework by replicating bills state-by-state. OpenAI spokesperson Liz Bourgeois defended the approach, saying it helps regulators enforce the law and gives the public clearer protections. Anthropic endorsed California's 2025 law regulating advanced AI models and has since backed stronger bills in New York, Illinois and Massachusetts. Fernandez said each bill was stronger than the previous one and moved real safety obligations forward, adding that transparency and self-reporting are no longer sufficient. He cited the company's Claude Mythos model, which was found capable of exploiting security flaws in major operating systems. OpenAI recently joined Anthropic in backing an Illinois law requiring annual independent third-party audits of safety plans. Anthropic has also endorsed a Massachusetts proposal it calls the nation's strongest, requiring independent evaluators to assess catastrophic risks and empowering the attorney general to enforce the mandate. Anthropic has also begun making direct political contributions to California legislators aligned with its AI safety views.