AI safety regulation strategy has diverged sharply between Anthropic and OpenAI as both companies lobby U.S. state legislatures with contrasting regulatory visions. Anthropic is actively endorsing and shaping increasingly stringent state-level AI safety laws, deliberately outpacing OpenAI’s push for uniform frameworks. Cesar Fernandez, Anthropic’s head of U.S. state and local government relations, stated the company aims to establish a progressively higher safety threshold for advanced models rather than merely supporting identical bills nationwide. This approach directly counters OpenAI’s reverse federalism campaign, led by lobbyist Chris Lehane, which seeks to build a national standard by replicating successful state legislation across multiple jurisdictions. Anthropic’s state-by-state strategy began with its endorsement of California’s 2025 advanced AI regulation, which it viewed as a stepping stone rather than a final benchmark. The company subsequently backed stricter proposals in New York, Illinois, and Massachusetts. The Illinois law, recently signed by Governor JB Pritzker, mandates annual independent third-party safety audits for leading AI developers. In Massachusetts, Anthropic supported even more rigorous provisions requiring independent catastrophic risk assessments and empowering the state attorney general to enforce compliance. Fernandez emphasized that transparency and self-reporting are no longer sufficient given the rapid evolution of capable models. The push for stricter controls stems from Anthropic’s internal safety assessments, particularly testing on its Claude Mythos model, which reportedly exploited vulnerabilities across major operating systems and prompted federal export controls during the Trump administration. Beyond legislative lobbying, Anthropic has intensified its political engagement, funding dueling super PAC networks and making direct contributions to California legislators who align with its safety-first ideology. Fernandez clarified that while the company does not direct employee donations, it leverages its workforce’s shared concerns to support aligned candidates, noting that its legislative strategy cannot wait for stalled federal action. OpenAI maintains that reverse federalism provides regulatory clarity, prevents conflicting compliance burdens, and enables developers to concentrate resources on safety. Spokesperson Liz Bourgeois noted that the company continues to evaluate state proposals, including the Massachusetts draft, while advocating for consistent safeguards. With federal lawmakers paralyzed and the White House adopting a fluctuating regulatory stance, state initiatives are becoming the primary venue for shaping U.S. AI policy. The divergent tactics of Anthropic and OpenAI will likely dictate whether American AI governance converges on a standardized framework or fragments into a cascading series of escalating state mandates.