Anthropic, on the other hand, saw the California law as a springboard to ratchet up its efforts on AI safety. Fernandez said the rapid development of increasingly powerful AI models was the main factor behind his company’s endorsement of more ambitious bills — in New York, Illinois and now Massachusetts — and its move to weigh in earlier in the legislative process.
“Each one of those bills was stronger than the previous bill, and the bills all moved real safety obligations forward,” Fernandez said. “Transparency and self-reporting, we don’t believe are sufficient anymore.”
He pointed to Anthropic’s powerful Claude Mythos model, which the company found to be capable of exploiting security flaws in every major computer operating system during its testing. The cybersecurity concerns raised by Mythos (and its public-facing version, known as Fable) sparked panic inside the Trump administration, which slapped export controls on the technology until Anthropic and the government could address alleged vulnerabilities.
Late last year, OpenAI lobbyists successfully pressed New York Gov. Kathy Hochul to amend her state’s AI safety bill to more closely resemble California’s rules. But to the surprise of some safety advocates, it joined Anthropic in backing an Illinois measure seen as stricter than those in New York and California. That proposal, signed into law this month by Gov. JB Pritzker, requires leading AI companies to submit to annual independent third-party audits of their safety plans — a first-of-its-kind mandate.
Anthropic is pushing the bar further. In late June, it endorsed regulations under development in Massachusetts for an economic development bond bill that Anthropic calls the nation’s strongest state AI safety proposal. The language it supported included a requirement for leading AI companies to hire independent evaluators to assess the potential for catastrophic risks such as the technology assisting in the development of bioweapons, as well as a provision empowering the state’s attorney general to enforce that mandate.
Bourgeois, the OpenAI spokesperson, said the company is still reviewing the Massachusetts proposal, but added OpenAI supports the state legislature’s focus on AI safeguards.
The AI giants have also clashed on the campaign trail. Each is associated with dueling super PAC networks that so far have sunk tens of millions of dollars into political campaigns across the country. And in June, Anthropic started cutting checks directly to California legislators.
“We back candidates for election and re-election when their point of view of AI safety regulation is aligned with our mission to make sure that the transition to a world with powerful AI does well for people in this country and throughout the world,” Fernandez said. “We’re very much supporting candidates where there’s ideological alignment.”
Fernandez said the company isn’t coordinating with employees who also have made contributions to political candidates in California and elsewhere.
“We don’t direct our employees to make contributions, but they work at Anthropic because they’re concerned about the future of AI and where this is headed if there’s not proper safety policy that’s enacted by governments,” said Fernandez. “I would assume that that drives them to engage in the political process.”