Anthropic pushes 3 states to toughen AI safety rules, outpacing OpenAI

Anthropic is urging states to pass progressively tougher AI safety laws rather than coalesce around a single rulebook, a strategy that pits the company directly against OpenAI's push for nationwide uniformity.

"While there are some in the industry that think of state policy as a way to create a ceiling for federal legislation, Anthropic is not just looking to support the same bill across the country in every single state," Cesar Fernandez, the company's head of US state and local government relations, said in an interview.

Anthropic was the only leading AI lab to endorse California's 2025 law, the first such statute in the country. It then backed an Illinois measure signed into law this month by Governor JB Pritzker that requires leading AI companies to submit to annual independent third-party audits of their safety plans — a first-of-its-kind mandate. In late June, Anthropic endorsed regulations under development in Massachusetts that would require independent evaluators to assess catastrophic risks such as the technology assisting in the development of bioweapons, with enforcement authority granted to the state's attorney general.

The split between the two AI giants comes as Congress remains paralyzed on AI legislation and the White House has shifted between light-touch and heavy-handed approaches. Whether state legislators coalesce around a single framework or compete to outdo each other will shape the final shape of US AI rules — and determine compliance costs for an industry projected to spend tens of billions of dollars on safety infrastructure.

The Divergent Strategies

OpenAI's top lobbyist, Chris Lehane, has coined the term "reverse federalism" to describe the company's attempt to bypass a paralyzed Congress and build a national AI framework by mirroring bills state-by-state. OpenAI spokesperson Liz Bourgeois defended the approach, saying "reverse federalism, where effective state safeguards shape national standards, helps regulators enforce the law, gives the public clearer protections, and allows developers to focus resources on safety rather than conflicting requirements."

Anthropic's Fernandez said the rapid development of increasingly powerful AI models drove the company's endorsement of more ambitious bills. "Each one of those bills was stronger than the previous bill, and the bills all moved real safety obligations forward," he said. "Transparency and self-reporting, we don't believe are sufficient anymore."

The Mythos Precedent

The stakes of the regulatory debate became clear in June, when Anthropic's Claude Mythos model — which the company found capable of exploiting security flaws in every major computer operating system — triggered a government intervention. The Trump administration forced Anthropic to take down Mythos and its safeguarded form, Fable 5, on June 12, citing a narrow jailbreak vulnerability. Anthropic disputed the action, calling the trigger a non-universal exploit. Access was restored on June 30 after an agreement with the government.

The episode galvanized the national security community and prompted new directives for addressing AI-enabled cyber threats. It also highlighted the absence of standard processes for government-industry collaboration on frontier AI risks, with participants describing it as a "pickup game" where rules were invented in the moment.

Political Spending and the Road Ahead

Anthropic has also begun cutting checks directly to California legislators since June, Fernandez confirmed. "We back candidates for election and re-election when their point of view of AI safety regulation is aligned with our mission," he said. The company is not coordinating with employees who have also made political contributions, he added.

OpenAI, meanwhile, successfully pressed New York Governor Kathy Hochul to amend her state's AI safety bill to more closely resemble California's rules. But it surprised some safety advocates by joining Anthropic in backing the stricter Illinois measure.

The AI giants are also associated with dueling super PAC networks that have sunk tens of millions of dollars into political campaigns across the country.

With no federal AI legislation in sight and the White House's position shifting, the state-by-state battle between the two approaches will determine whether the US adopts a patchwork of escalating safety requirements or a harmonized national standard — with billions of dollars in compliance costs and competitive advantage at stake.

This article is for informational purposes only and does not constitute investment advice.