According to an article by Maxwell Zeff published in WIRED, artificial intelligence company Anthropic is leading a major political push aimed at accelerating and tightening regulation of AI models at the state level in the United States. The company, currently valued at nearly $1 trillion, expressed support last year for the first wave of safety and transparency laws for frontier AI models in California and New York. These laws were fiercely opposed by large segments of Silicon Valley, where critics argued they would stifle the momentum of artificial intelligence development. Now, however, Anthropic argues that these laws may already be outdated, and it is pushing states to adopt even tougher regulations.
Cesar Fernandez, Anthropic’s head of US state and local government relations, explained in an interview with WIRED that "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." Fernandez added that in the company's view, "transparency and self reporting are no longer sufficient safety measures for the most powerful AI systems."
Anthropic's Safety Vision and Strict Laws in Illinois and Massachusetts
For a startup with such a high valuation, an advocacy position supporting tight regulation might seem unusual, but Anthropic operates under a unique model. As previously reported in WIRED, Anthropic's leaders believe they must build a massive business based on developing and selling access to advanced artificial intelligence in order to fulfill their founding mission: ensuring that the global transition to transformative AI is carried out safely.
As the company has grown, it has begun endorsing some of the harshest proposed regulations in the United States targeting developers of advanced AI models. These rules are designed to mitigate catastrophic risks, including the possibility that advanced models could assist in creating financial disasters or events resulting in mass casualties. Beyond the self-reporting laws in California and New York, Anthropic supported a bill in Illinois that would require AI labs to submit their safety processes to evaluations by third-party auditors.
Recently, the company also expressed support for a bill in Massachusetts. In addition to requiring third-party audits of AI labs, this legislation would grant the state's attorney general the authority to seek injunctive relief against companies that fail to comply with safety requirements.
Cesar Fernandez's Expertise and the Paralysis in Congress
Fernandez, who joined Anthropic earlier this year, brings extensive experience in local policy battles. He previously served as head of state government relations at sports betting giant FanDuel and as a senior public policy associate at Uber, helping both companies navigate and win regulatory battles in various states across the US.
His expertise has become particularly crucial for Anthropic, given that the US Congress continues to stall and has failed to pass comprehensive federal legislation regulating artificial intelligence. As a result of this legislative paralysis in Washington, individual US states are taking the lead and shaping the legislative landscape in this field.
Criticism in Silicon Valley: Accusations of "Regulatory Capture"
While Anthropic's pro-regulation agenda has garnered support from AI safety groups, labor unions, and company allies, some figures in Silicon Valley interpret its political efforts in a negative light, accusing the company of harboring ulterior motives.
David Sacks, the former White House AI czar and current technology adviser to President Donald Trump, argued that Anthropic is attempting to pass cumbersome laws to trap small startups in bureaucratic red tape, thereby securing its leading position in the AI race and preventing competition. "Anthropic is running a sophisticated regulatory capture strategy based on fear-mongering," Sacks wrote in a social media post last year. According to him, the company is the primary driver behind the state-level regulatory frenzy that is damaging the startup ecosystem.
Anthropic's Defense: Revenue Threshold and High Development Costs
Fernandez completely denies Sacks' accusations. He notes that Anthropic has exclusively supported state-level bills that target "large AI model developers." While this definition is formulated differently in each proposed bill, it generally refers to companies that have spent hundreds of millions of dollars on AI development and whose annual revenues exceed $500 million. "It’s hard to imagine a startup meeting that threshold," Fernandez remarked in the interview.
However, in a world where developing a competitive, advanced AI model requires raising billions of dollars in capital, there are highly promising startups that could potentially reach these thresholds in the foreseeable future. Companies like Safe Superintelligence, Thinking Machines Lab, and Mistral have all raised billions of dollars from investors, even though their current revenues remain significantly lower than those of giants like Anthropic or OpenAI. While these are not typical startups, they certainly represent potential competitors to Anthropic.
On the other hand, Fernandez's primary counterargument is that any company large enough to develop a powerful AI model should be subject to the same regulations, as all such models share the same fundamental risks. He noted that part of Anthropic's objective is to "inspire a race to the top developing the most safe and secure AI systems," a mission that includes pushing for government legislation. "Our mission is to make sure that this transition to a world with advanced AI goes well for Americans and humanity," Fernandez said.
The Regulatory Boundary: State Powers vs. Federal Government
Despite its efforts at the state level, Anthropic draws clear red lines regarding what states should be permitted to do in the absence of federal legislation. In a policy document published by the company last month, it recommended that governments maintain a mechanism that allows them to block companies from deploying new AI models if they are determined to be unsafe.
This recommendation appears somewhat ironic, considering that the Trump administration recently ordered Anthropic to suspend access to its two most powerful models—Mythos and Fable 5—for foreign nationals, a directive the company did not welcome. Anthropic believes that the authority to block the deployment of unsafe AI models should be reserved exclusively for the federal government, rather than state-level legislators, though Fernandez noted that this remains an evolving conversation.
After the Trump administration sent the export control directive that forced the company to take its Mythos and Fable 5 models offline for all users, Anthropic published a blog post arguing that blocking model deployments should only be executed through a fair and transparent evaluation process.
Accusations Against Alibaba and the Open Source Controversy
Fernandez declined to elaborate on the company's federal policy initiatives, which do not fall under his purview, though the company is highly active in Washington as well. Last month, Anthropic sent a letter to the US government accusing Chinese tech giant Alibaba of conducting a "distillation attack"—systematically extracting information from Anthropic's models using sophisticated prompts to develop its own AI tools.
Some AI researchers dismissed these allegations, viewing them as another attempt at regulatory capture. These researchers argued that Anthropic's real objective is to pressure the US government to ban Chinese open-weight models, a step that could force thousands of American businesses relying on those models to transition to Anthropic's enterprise offerings. Anthropic CEO Dario Amodei has previously warned Congress about the dangers associated with open-source artificial intelligence. At the state level, Fernandez clarified that Anthropic has not specifically targeted open-source models in the bills it has supported, emphasizing instead that the key factors are a model's capabilities and its risks.
Legislative Priorities: Public Concerns vs. Catastrophic Risks
The concerns most frequently raised by American voters—such as potential job losses due to technology, the environmental and community impacts of constructing new data centers, and the effects of chatbots on children—have not driven a comparable legislative campaign from major AI labs.
While Anthropic and several of its competitors have promised that ordinary taxpayers will not be forced to pay for data centers, and Anthropic has published proposals to address future AI-driven job displacement, the industry has spent significantly less political capital on state-level laws addressing these everyday issues.
When asked about this discrepancy in priorities, Fernandez replied that Anthropic is "eager to engage with lawmakers" on various issues beyond catastrophic risks, and that it is already holding such discussions in various states. However, as of now, these conversations have not translated into a coordinated and organized state-level effort similar to the one Anthropic is leading regarding the safety and existential risks of frontier models.