Anthropic, the AI safety company valued at nearly $1 trillion, is urging US states to go beyond current transparency laws and adopt stricter regulations for the most powerful AI systems, arguing that self-reporting is no longer enough.
The company, which supported the first wave of frontier AI safety laws in California and New York last year, now says those transparency-focused bills passed in 2025 are already outdated. Much of Silicon Valley fought against those earlier laws, arguing they would stifle the AI boom. But Anthropic is now pushing for a new generation of rules that require third-party auditing and give states the power to block unsafe deployments.
“The transparency-focused safety bills of 2025 were a really important start, but as the capabilities of AI systems continue to advance quickly—the policy responses need to match,” said Cesar Fernandez, head of US state and local government relations at Anthropic, in an interview. “We think that transparency and self reporting are no longer sufficient safety measures for the most powerful AI systems.”
A new push for third-party auditing
Anthropic’s founding mission is “to ensure that the world safely makes the transition through transformative AI,” according to the company. To that end, it has endorsed an Illinois measure requiring AI labs to get their safety processes evaluated by third-party auditors. It also backed a Massachusetts policy that mandates third-party auditing for AI labs and empowers the state attorney general to seek injunctive relief from non-compliant companies.
Fernandez, who joined Anthropic earlier this year after serving as head of US state government relations for FanDuel and as a senior public policy associate at Uber, said the company believes these measures are necessary as AI capabilities accelerate. Congress continues to stall on passing federal AI regulation, leaving states to take the lead.
“If you're an AI model developer that's developing powerful AI systems, there are requirements to be transparent with people on what you're building, the model capabilities, and the risks associated with that model to critical infrastructure in states—all of that should be top priority,” Fernandez said.
The company published a policy document last month recommending that governments have a mechanism to block companies from deploying unsafe AI models. Anthropic argues that such blocking powers should be reserved for the federal government, not state lawmakers, to ensure a fair and transparent evaluation process. The document outlines a vision where a federal agency would evaluate AI models before they reach the public, a process Anthropic says would prevent a patchwork of state-level bans.
Anthropic’s push comes as AI capabilities accelerate rapidly. The company’s latest model releases brought the cybersecurity capabilities of advanced AI into the national spotlight. Concerns raised more often by American voters include losing jobs to AI, the negative impacts of data centers, and the effects of chatbots on children. Anthropic and its rivals have promised that ordinary taxpayers won’t be stuck paying for data centers. The company has also published proposals for responding to future AI-driven job displacement.
Accusations of regulatory capture
Not everyone is convinced by Anthropic’s pro-regulation agenda. David Sacks, the former White House AI czar and technology adviser to President Donald Trump, has been a vocal critic. Last year, Sacks wrote on social media: “Anthropic is running a sophisticated regulatory capture strategy based on fear-mongering.” He added, “It is principally responsible for the state regulatory frenzy that is damaging the startup ecosystem.”
Sacks and other critics argue that Anthropic is trying to get cumbersome laws passed to trap smaller AI startups in red tape, securing its own dominant position. Fernandez outrightly denies the accusation. He says Anthropic only supports state AI bills that apply to “large AI model developers.”
The definition of “large AI model developers” varies per bill, but generally refers to companies that spent hundreds of millions of dollars on AI development and have more than $500 million in annual revenue. “It’s hard to imagine a startup meeting that threshold,” Fernandez said.
Potential competitors like Safe Superintelligence, Thinking Machines Lab, and Mistral have each raised billions from investors, but their revenue remains far lower than Anthropic or OpenAI. That means they would likely be exempt from the regulations Anthropic is endorsing. Critics say this creates a moat that protects established players while burdening newcomers who might challenge the status quo.
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
A race to the top or a moat?
Fernandez says part of Anthropic’s goal is to “inspire a race to the top developing the most safe and secure AI systems.” The company published a policy document last month recommending that governments have a mechanism to block companies from deploying unsafe AI models.
But the suggestion is ironic given Anthropic’s reaction to the Trump administration’s recent directive. The administration told Anthropic to suspend access to its two most powerful AI models for foreign nationals. Anthropic wasn’t a fan of the move. After the export control directive, Anthropic took its Mythos and Fable 5 models offline for everyone.
Anthropic wrote in a blog post that blocking AI model deployments should only happen through a fair, transparent evaluation process. The company thinks the right to block unsafe AI deployments should be reserved for the federal government, not state lawmakers.
Last month, Anthropic sent a letter to the US government accusing Chinese tech giant Alibaba of a distillation attack. Some AI researchers dismissed Anthropic’s claims as “regulatory capture by another means.” Critics argue Anthropic’s real goal is to persuade the US government to ban open-weight Chinese models, driving businesses to Anthropic’s enterprise offerings.
Anthropic CEO Dario Amodei has warned Congress about the dangers of open source AI in the past. But Fernandez says Anthropic hasn’t specifically targeted open source AI models in any legislation it has endorsed. “It’s less of a question of the model construction and more of the model capabilities, and when they rise to a certain level, they should be captured in these state frameworks,” he said.
Beyond existential risks
Anthropic has spent years warning lawmakers about the catastrophic risks of advanced AI. The company’s latest model releases brought the cybersecurity capabilities of advanced AI into the national spotlight. But concerns raised more often by American voters include losing jobs to AI, the negative impacts of data centers, and the effects of chatbots on children.
Anthropic and its rivals have promised that ordinary taxpayers won’t be stuck paying for data centers. The company has also published proposals for responding to future AI-driven job displacement. Yet the industry has put considerably less political muscle behind state laws addressing job loss, data centers, and children’s issues compared to existential risks.
Fernandez says Anthropic is “eager to engage with lawmakers” on issues beyond catastrophic AI safety risks and is having those conversations throughout states. “Our mission is to make sure that this transition to a world with advanced AI goes well for Americans and humanity,” he said.
But those discussions have not produced a coordinated state-level push like the one Anthropic has mounted around existential risks. For now, the company’s focus remains on the most powerful systems—and on shaping the rules that will govern them.
The debate over Anthropic’s motives is unlikely to fade. Supporters see a company living up to its safety-first mission. Skeptics see a strategic play to lock in advantages while competitors struggle to catch up. Either way, the outcome of these state-level battles will shape the AI industry for years to come.