Anthropic, a leading artificial intelligence research company, is now actively pushing for more stringent AI safety regulations at the state level, a significant evolution from its earlier support for transparency-focused laws. This shift comes as the company acknowledges the rapidly advancing capabilities of AI systems, arguing that current policy responses may already be falling behind. While initially instrumental in securing new transparency requirements in California and New York last year – measures opposed by much of Silicon Valley on grounds of stifling innovation – Anthropic now believes these initial steps are insufficient for the most powerful AI models.
Cesar Fernandez, Anthropic’s head of US state and local government relations, articulated this evolving stance in a recent interview. "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 stated. "We think that transparency and self-reporting are no longer sufficient safety measures for the most powerful AI systems." This proactive approach to regulation, particularly from a startup now valued at close to $1 trillion, underscores Anthropic’s unique founding mission: "to ensure that the world safely makes the transition through transformative AI."
This commitment to safety is deeply embedded in Anthropic’s operational philosophy. Founded with the explicit goal of developing and commercializing advanced AI while prioritizing safety, the company has consistently articulated a vision where massive business growth is intrinsically linked to responsible AI development. This dual objective places Anthropic in a distinctive position within the AI landscape, often at odds with industry peers who prioritize rapid development and market expansion with less emphasis on regulatory oversight.
The company’s growing advocacy for stringent regulations targets potential catastrophic AI risks, including scenarios involving financial instability and mass casualties. Beyond the initial transparency mandates in California and New York, Anthropic has thrown its weight behind measures requiring third-party audits of AI safety processes, such as an Illinois initiative. More recently, the company has endorsed a Massachusetts policy that not only mandates third-party auditing but also empowers the state’s attorney general to seek injunctive relief against non-compliant AI firms. Fernandez, who joined Anthropic this year after significant experience in government relations for major tech and sports betting companies like FanDuel and Uber, is now leveraging his expertise to navigate this complex regulatory terrain, especially as federal AI legislation continues to stall.
The Regulatory Capture Debate: Accusations and Rebuttals
Anthropic’s increasingly assertive regulatory agenda has drawn both praise and sharp criticism. While AI safety advocates, labor unions, and allied organizations commend the company’s proactive stance, some figures within the tech industry view its lobbying efforts with suspicion. David Sacks, a prominent venture capitalist and former White House AI czar, has publicly accused Anthropic of employing a "sophisticated regulatory capture strategy based on fear-mongering." Sacks argues that Anthropic aims to inundate smaller AI startups with burdensome regulations, thereby cementing its own leadership position in the AI race and hindering emerging competitors.
"Anthropic is running a sophisticated regulatory capture strategy based on fear-mongering," Sacks stated in a social media post last year. "It is principally responsible for the state regulatory frenzy that is damaging the startup ecosystem."
Fernandez firmly refutes these allegations, emphasizing that Anthropic’s supported legislation is specifically designed to apply to "large AI model developers." He clarified that this designation typically refers to companies with substantial AI development investments, often in the hundreds of millions of dollars, and significant annual revenues exceeding $500 million. "It’s hard to imagine a startup meeting that threshold," Fernandez commented.
However, the financial landscape of frontier AI development is rapidly evolving. The immense capital required to create competitive AI models means that a select few startups could soon meet or exceed these thresholds. Companies like Safe Superintelligence, Thinking Machines Lab, and Mistral, which have collectively raised billions, represent potential future competitors. While their current revenues may lag behind giants like Anthropic or OpenAI, their substantial funding positions them as significant players in the long term.
From Anthropic’s perspective, the argument for regulation is grounded in the principle that any entity capable of developing powerful AI models should be subject to commensurate safety standards due to the inherent risks associated with the technology. Fernandez articulated this rationale, stating that Anthropic’s objective is to "inspire a race to the top developing the most safe and secure AI systems." This, he contends, necessitates advocating for robust government oversight. "Our mission is to make sure that this transition to a world with advanced AI goes well for Americans and humanity," Fernandez asserted. "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."
Defining the Boundaries: Federal vs. State Authority
While Anthropic champions increased state-level regulation, it also draws a line regarding the extent of authority state governments should wield in the absence of comprehensive federal AI policy. In a recent policy document, Anthropic proposed that governments should possess mechanisms to block the deployment of new AI models deemed unsafe. This suggestion carries a degree of irony, given recent directives from the Trump administration compelling Anthropic to suspend access to its most advanced AI models for foreign nationals, a move the company did not welcome.
Anthropic’s position is that the ultimate authority to block the deployment of unsafe AI models should reside with the federal government, not individual state legislatures. Fernandez acknowledged that this is an evolving discussion. Following the US government’s export control directive that led to the temporary offline status of Anthropic’s Mythos and Fable 5 models, the company articulated in a blog post that such actions should only occur through a fair and transparent evaluation process.
Although Fernandez refrained from detailing Anthropic’s federal policy engagements, the company has been actively involved in discussions at the national level. Last month, Anthropic submitted a letter to the US government alleging that Chinese tech giant Alibaba engaged in a "distillation attack" – systematically prompting Anthropic’s models to extract information for its own AI development. This claim has been met with skepticism by some AI researchers who suggest it might represent another form of regulatory capture, aiming to persuade the US government to ban open-weight Chinese models. Such a ban could potentially steer American businesses reliant on these models toward Anthropic’s enterprise offerings. This aligns with previous warnings from Anthropic CEO Dario Amodei to Congress regarding the potential risks associated with open-source AI.
At the state level, however, Fernandez maintains that Anthropic has not specifically targeted open-source AI models in its legislative endorsements. "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 explained.
Broader Implications and Industry Focus
Regardless of the perceived motivations behind Anthropic’s regulatory advocacy, the company is undeniably playing a significant role in shaping the future of AI policy. Its recent model releases have amplified national attention on the cybersecurity implications of advanced AI. Prior to this, Anthropic had spent years alerting lawmakers nationwide to the potential catastrophic risks posed by frontier AI systems.
This focus on existential risks stands in contrast to the AI concerns more frequently voiced by the general American public, such as job displacement, the environmental impact of data centers, and the effects of AI chatbots on children. While Anthropic and its competitors have pledged to mitigate the financial burdens of data center expansion on taxpayers and have published proposals addressing AI-driven job displacement, the industry has historically dedicated less political capital to state-level legislation addressing these more immediate societal concerns.
When questioned about this disparity, Fernandez stated that Anthropic is "eager to engage with lawmakers" on issues beyond catastrophic AI safety risks and is "having those conversations throughout the states." However, current efforts have not yielded the same coordinated state-level push seen concerning the risks associated with frontier AI models. This suggests a continued prioritization of high-stakes, potentially civilization-altering risks by leading AI labs, while more common public concerns may receive comparatively less industry-backed legislative attention.
The ongoing dialogue between AI developers, policymakers, and the public will be crucial in determining the trajectory of AI regulation. Anthropic’s evolving position highlights the dynamic nature of the field and the increasing pressure on companies to balance innovation with a robust commitment to safety and societal well-being.
This article is an edition of Maxwell Zeff’s Model Behavior newsletter. Read previous newsletters here.
