How Trump's Anthropic Move Is Testing The Legal Limits Of Tech Restrictions

President Donald Trump’s action cutting off foreign access to Anthropic’s most powerful AI models is stretching the legal limits of government control on tech exports. It’s a warning for other AI companies.

The administration is using export controls — rules that restrict the transfer of sensitive technologies to foreign parties — to bar Anthropic from allowing its Fable 5 and Mythos 5 models to be used by foreign nationals, including those inside the United States, without a license. The directive prompted the company to disable access to the service on Friday.

The Commerce Department, which oversees export control rules, has only applied those restrictions to Anthropic’s models, and has not formalized the rule by publishing it in the Federal Register. But if left unchallenged, the maneuver could embolden the agency to impose the same restrictions on high-end models across the entire AI industry, allowing them to potentially choke off access for any foreign person who uses models such as OpenAI's ChatGPT and Google's Gemini.

“If this is going to be their position going forward with every other model or every other data center, that will be a dramatic shift,” said Kevin Wolf, a former senior Commerce Department official who helped craft export control measures during the Obama administration.

The decision to slap export controls on Anthropic rests on authorities granted under the Export Control Reform Act of 2018, which established a formal licensing system requiring American companies to seek government approval before transferring a broad range of sensitive technologies to foreign parties. The law gave Commerce sweeping authority to designate which technologies require a license, a list that has since grown to include certain AI-related software and systems.

But it’s not clear if the law gives the federal government the authority to regulate whether a foreign person can simply log in and use an American AI model.

“This is the first time I think I'm aware of — and I was at [the Commerce Department bureau overseeing export controls] for 25 plus years — where we tried to restrict access to a thing that was unrelated to access to the underlying technology,” said Matthew Borman, a former deputy assistant secretary of Commerce, who served during both the first and second Trump administrations.

Borman said other AI developers should take note — particularly those who have yet to engage with the government on their models.

Trump unveiled an executive order earlier this month asking AI companies to voluntarily submit their models to the government 30 days before they plan to release them, and the White House action on Anthropic is likely to encourage other companies to follow suit.

"If they don't, they could potentially run the risk that they go ahead and roll something out, and the government sends them a cease-and-desist directive," Borman said. "I think this Anthropic issue is quite interesting for them."

Borman added that the department's own guidance cuts against the decision to cover foreign persons.

He pointed to previous guidance from the Commerce Department’s Bureau of Industry and Security that concluded simply allowing a foreign national to access cloud computing capability did not itself constitute an export. The directive to Anthropic, he said, appears to conflict with that guidance.

Wolf was also skeptical. “I don’t know what the legal authority is for that,” he said.

Anthropic, however, has not given any indication it plans to sue, chalking up the designation to a "misunderstanding" and denying that the security vulnerability cited by the government warranted pulling the models entirely.

The company dispatched senior staff to Washington for a meeting Monday with Commerce officials and the National Cyber Director's office in hopes of removing the new restrictions on its AI models. That meeting ended without a resolution.

A representative for Anthropic did not immediately respond to a request for comment.

Companies like Open AI are “very clearly on the trajectory to be as capable as Mythos or Fable in the not too distant future,” said Alasdair Phillips-Robins, a fellow at the D.C. think tank, Carnegie Endowment for International Peace who worked on former President Joe Biden’s export control policies while at the Commerce Department. “It's become clear that the White House is going to, very plausibly, take that seriously, and may do some very aggressive things in response,” he said.

An OpenAI official who briefed reporters on Wednesday after an AI lunch at the G7 Summit in France pointed to Trump’s June 2 executive order when asked whether the company is worried it could also be the target of U.S. export controls. It's “incredibly important” that companies comply with certain government protocols, said the official, who was granted anonymity per the ground rules of the briefing.

The person added: “I also think, as we go forward, and these models become more and more capable, the muscles and these frameworks are going to continue to need to evolve and grow.”