Anthropic Wants Stricter AI Laws. Not Everyone Thinks Its Motives Are Pure.
The maker of the Claude chatbot is pushing states to regulate frontier AI harder than anyone in Silicon Valley expected. The company says safety demands it. Critics say it is a business play dressed up as public interest.
Key points
- Anthropic, valued at close to $1 trillion, is actively lobbying US states to pass stricter rules on the most powerful AI systems.
- California and New York passed transparency laws in 2025 that Anthropic supported, but the company now says those laws do not go far enough.
- Anthropic has endorsed an Illinois measure and a Massachusetts bill that would require AI companies to hire outside auditors to inspect their safety practices.
- The bills Anthropic backs apply only to companies with over $500 million in annual revenue that have spent hundreds of millions building AI, which currently covers a small group of large firms.
- Former White House AI adviser David Sacks has accused Anthropic of using regulation to kneecap smaller rivals, a charge Anthropic flatly denies.
A company worth nearly $1 trillion lobbying for tougher rules on itself sounds strange. Anthropic is doing exactly that.
The San Francisco company, which makes the Claude AI assistant, has been working with state governments across America to advance some of the strictest proposed rules on frontier AI, meaning the most powerful AI systems that exist today. Anthropic's head of US state and local government relations, Cesar Fernandez, confirmed the push in a recent interview with Wired AI.
"Transparency and self-reporting are no longer sufficient safety measures for the most powerful AI systems," Fernandez said.
Anthropicbacked new transparency laws in California and New York in 2025, which most of Silicon Valley fought hard against. Now it says those are already outdated. It has backed an Illinois bill requiring AI labs to bring in third-party auditors, independent outside firms that check whether a company's safety practices actually work. Most recently it endorsed a similar Massachusetts proposal that would also give the state's attorney general power to take AI companies to court if they refuse to comply.
Are these rules just designed to hurt Anthropic's competitors?
That is the accusation, and it comes from a credible voice. David Sacks, who served as the Trump administration's AI adviser, wrote publicly last year that Anthropic is "running a sophisticated regulatory capture strategy," meaning using government rules as a weapon to protect its own market position rather than to protect the public. We have covered Sacks's interventions in AI policy in four stories since July.
Fernandez denies it. He points out that every bill Anthropic has backed applies only to companies with more than $500 million in annual revenue that have already spent hundreds of millions developing AI. "It's hard to imagine a startup meeting that threshold," he says.
That defence has a weak spot. Several well-funded newer AI companies, including Safe Superintelligence and Mistral, have each raised billions of dollars and could reach those thresholds. They are also direct competitors to Anthropic.
The stronger version of Anthropic's argument is simpler: if a company is big enough to build a genuinely powerful AI system, it is big enough to cause serious harm. Same risk, same rules. Whether that logic holds up is precisely what auditors would be paid to decide, and as our reporting on 16 July found, proving that safety practices hold under scrutiny is much harder than testing them in the first place.
Anthropics federal lobbying is also raising eyebrows. Last month the company sent a letter to the US government accusing Chinese tech giant Alibaba of a "distillation attack," claiming Alibaba systematically questioned Anthropic's models to extract knowledge and build its own tools. Some AI researchers say that accusation is itself regulatory capture by another route, an attempt to persuade Washington to block open-weight Chinese AI models (models whose underlying code is publicly available) and steer American businesses toward Anthropic instead.
If you run a business that uses AI tools, the practical implication is this: watch your state legislature now. Auditing requirements and liability rules are moving faster than most companies expect, and they will shape which AI products survive in the US market over the next two to three years. Getting ahead of that costs nothing; scrambling to comply after the fact usually does.
