Anthropic Says Last Year's AI Safety Laws Are Already Too Weak, Pushes States for Tougher Rules

A trillion-dollar company wants tougher rules on itself

Anthropic spent 2025 pushing California and New York to pass transparency and self-reporting laws for frontier AI companies, according to Wired. Much of Silicon Valley fought those laws, arguing they'd slow down the AI boom. Anthropic backed them anyway.

Now the company says that wasn't enough.

Cesar Fernandez, Anthropic's head of US state and local government relations, told Wired the 2025 laws were "a really important start" but that AI capabilities are moving faster than the policy response. "We think that transparency and self reporting are no longer sufficient safety measures for the most powerful AI systems," Fernandez said.

So Anthropic is now backing a new round of state proposals with real teeth. In Illinois, the company supports a measure requiring AI labs to submit their safety processes to third-party auditors, not just self-report. In Massachusetts, Anthropic endorsed a policy that goes further still: mandatory third-party audits plus the power for the state attorney general to seek injunctive relief against companies that don't comply, according to Wired.

That's a real legal hammer. An attorney general seeking an injunction isn't a slap on the wrist. It's a court order that can shut down a product or a process.

Why a company would lobby to regulate itself harder

On its face, this looks backwards. Companies almost never ask governments to make their own lives harder. Fernandez joined Anthropic earlier this year after running state government relations for FanDuel and working in public policy at Uber, according to Wired, two companies that spent years fighting regulation, not inviting it.

Anthropic's position only makes sense if you take its founding mission seriously. The company says it exists "to ensure that the world safely makes the transition through transformative AI," as Wired describes it. Anthropic's leadership has argued publicly, including in prior Wired reporting, that the business exists to fund that safety mission, meaning the regulatory push is at least consistent with the story Anthropic tells about itself.

There's also a more cynical, and equally plausible, read. Anthropic is now valued at nearly $1 trillion. A company that size can absorb the cost of third-party audits and compliance staff far more easily than a startup with twelve employees trying to build a competing model. Tougher regulation locks in the current leaders and raises the price of admission for everyone else. That's not a conspiracy theory, it's just how regulatory capture usually works in any industry, from banking to pharmaceuticals.

Sacks calls it self-serving

David Sacks, the White House AI czar and technology adviser to President Trump, has pushed back hard on Anthropic's motives, per Wired's reporting. He has characterized the company's regulatory push as less about safety and more about strategy, according to Wired's reporting of his public argument.

Sacks's skepticism deserves a fair hearing. If tougher audit-and-injunction regimes make it structurally harder for smaller AI labs to compete, that's a legitimate competitive-policy concern, not just partisan noise. Regulatory barriers that look like safety measures but functionally protect incumbents are a well-documented pattern in industries from telecom to finance. Nobody should wave that concern away just because it comes from a Trump adviser.

At the same time, AI safety groups and labor unions have praised Anthropic's approach, according to Wired. The company has been consistent since 2025: it backed the original California and New York laws even when most of Silicon Valley opposed them. That's a track record, not just a talking point.

What's actually unresolved

Congress has not passed federal AI legislation, according to Wired, leaving states to write the rules piecemeal. Illinois and Massachusetts lawmakers are now deciding whether to hand their attorneys general injunction power over AI labs, a decision that will shape how enforcement actually works if these bills pass.

Neither the Illinois nor Massachusetts measure has become law as of this writing. Whether state legislators adopt Anthropic's preferred audit-and-enforcement model, or whether competing AI companies successfully lobby against it as anti-competitive, will determine whether this becomes the template other states copy or a case study in how a dominant firm shaped rules in its own favor.