Artificial intelligence giant Anthropic is pursuing a strategy of one-upmanship that encourages countries to increasingly tighten AI guardrails rather than aligning on a single regulation.
This approach stands in stark contrast to that espoused by the company’s biggest rival, OpenAI, which is pushing state legislatures toward common ground on regulating breakthrough technologies.
“While some in the industry see state policies as a way to cap federal law, Anthropic is not simply trying to support the same legislation in every state across the country,” Cesar Fernandez, the company’s head of U.S. state and local government, said in an interview with Politico (which, like Business Insider, is part of the Axel Springer Global Reporters Network) on Tuesday. “We are looking for legislation that meaningfully raises the bar for safety for our most capable AI systems.”
Fernandez’s comments were in response to POLITICO’s questions about OpenAI’s ongoing campaign to shape state AI regulations. ChatGPT creator and top lobbyist Chris Lehane coined the term “reverse federalism” to describe attempts to bypass a paralyzed Congress and create a national AI framework that reflects state-by-state legislation.
The veiled attack on OpenAI is on-brand for Anthropic, whose executives left OpenAI in 2020 over concerns that the company wasn’t prioritizing safety. Anthropic has consistently pushed for stronger AI safety rules at both the federal and state levels, but some critics, particularly those close to the Trump administration and venture capital insiders, have criticized the efforts as an attempt to thwart regulators and lock out competitors.
In a statement, OpenAI spokeswoman Liz Bourgeois defended her company’s approach as “reverse federalism, where effective national safeguards shape national standards, help regulators enforce the law, provide clearer protections for the public, and allow developers to focus resources on safety rather than conflicting requirements.”
The split in OpenAI and Anthropic’s approach to state legislatures comes at a critical time for AI regulation. With Congress reluctant to act and the White House vacillating between lenient and harsh responses, the AI industry is increasingly looking to states for regulatory clarity. Whether state legislatures ultimately unite around a single AI safety framework or strive to outdo each other over time will have a major impact on the final shape of AI regulation in the United States.
Like Lehane, Fernandez wants a federal framework, but said the government’s response to the risks posed by advanced AI models “cannot wait for Washington to act.”
The human lobbyist also distinguished his company by touting its early entry into national policy discussions. Anthropic is the only major AI institute to support California’s 2025 law regulating advanced AI models, the first such law in the country.
OpenAI did not take a position on California’s proposal prior to its passage. But the government has since looked to the law, which aims to make companies’ safety plans more transparent, as an example for other states to emulate.
Meanwhile, Anthropic saw the California law as a starting point to strengthen its AI safety efforts. Fernandez said the rapid development of increasingly powerful AI models was a key factor in the company’s move to support more ambitious bills in New York, Illinois, and now Massachusetts, moving them earlier in the legislative process.
“Each of these bills is stronger than the previous one, and all of them advance substantial safety mandates,” Fernandez said. “We believe that transparency and self-reporting are no longer enough.”
He pointed to Anthropic’s powerful Claude Mythos model. This model was found by the company to be capable of exploiting security flaws in all major computer operating systems during testing. Cybersecurity concerns raised by Mythos (and its public version known as Fable) caused panic within the Trump administration, which imposed export controls on the technology until Anthropic and the government could address the alleged vulnerabilities.
Late last year, OpenAI lobbyists successfully pressured New York Governor Kathy Hochul to amend her state’s AI safety bill to more closely align with California’s rules. But to the surprise of some safety advocates, the company joined Anthropic in supporting Illinois’ measures, which are said to be stricter than New York and California. The proposal was signed into law by Gov. J.B. Pritzker this month. This is the first of its kind to require major AI companies to submit annual audits of their safety plans to independent third parties.
Anthropic is pushing the envelope even further. In late June, Anthropic endorsed regulations being developed in Massachusetts for an economic development bond bill that Anthropic called the nation’s strongest national AI security proposal. Supported language included a requirement for large AI companies to hire independent evaluators to assess the potential for catastrophic risks, such as technologies that support the development of biological weapons, and a provision that would authorize state attorneys general to enforce that obligation.
Bourgeois, the OpenAI spokesperson, said the company is still considering Massachusetts’ proposal, but added that OpenAI supports the state Legislature’s focus on AI safeguards.
AI giants also clashed during election campaigns. Each is involved in a dueling battle with super PAC networks that have pumped tens of millions of dollars into political campaigns across the country. And in June, Anthropic began writing checks directly to California state legislators.
“We will support candidates for election and re-election if their perspectives on AI safety regulation align with our mission to ensure the transition to a world with strong AI works for people in this country and around the world,” Fernandez said. “We strongly support candidates with whom we align ideologically.”
Fernandez said the company does not coordinate with employees who have donated to political candidates in California and elsewhere.
“While we do not instruct our employees to donate, our employees work at Anthropic because they are concerned about the future of AI and where it will go without proper safety policies enacted by the government,” Fernandez said. “I think that’s what motivates them to participate in the political process.”
The Axel Springer Global Reporter Network leverages the resources of its newsroom to publish ambitious scoops, investigations, interviews, opinion pieces and analysis. This allows journalists from POLITICO, Business Insider, WELT, BILD, Onet, Fakt and more to collaborate on important stories that reach an international audience of hundreds of millions across platforms including online, print, TV and audio.