Anthropic sends top security experts to Washington to rescue flagship models

Following a report by Cryptopolitan regarding an imposed federal export ban on the sale of its AI software to users who are not Americans, Anthropic has deployed security researchers and AI models to Washington.

The Monday meetings were the first face-to-face talks between the startup and senior Trump administration officials since the order took effect.

The White House does not expect the dispute to disappear within a few days, though officials have not ruled out a fast settlement.

According to Politico, one senior official placed the burden on the company, saying, “That’s up to Anthropic.” The restriction was introduced over possible security weaknesses in the new model and now blocks non-U.S. users from reaching the company’s top technology.

Anthropic sends its security team to explain the model’s defenses

Anthropic spent the weekend trying to keep the issue from growing, as co-founder Tom Brown and Sarah Heck, who leads public policy at the company, joined long calls with Commerce Secretary Howard Lutnick and National Cyber Director Sean Cairncross.

These communications led to meetings on Monday involving the Commerce Department and the office of Cairncross. The staff took care of the discussions. One such official was Chris Fall who heads the Center for AI Standards and Innovation at the Commerce Department.

At the meeting, Anthropic presented the security measures surrounding the model in great detail to the federal employees. It talked about testing systems, looking for vulnerabilities, and limiting undesirable behavior before releasing the product. This was aimed at convincing Washington that it could offer the model overseas without presenting any national security threat.

The company sent Logan Graham from the Frontier Red Team, where he tests systems by trying to break their protections. Dave Orr, the head of safeguards, joined him. Nicholas Carlini, the company’s lead security researcher, was also part of the delegation.

The government action has already removed foreign access to Anthropic’s strongest models. It has also created uncertainty over whether federal officials can stop an AI release after a company follows a voluntary review system.

Trump officials defend the order while industry leaders warn about wider damage

Trump’s latest AI executive order has requested that major tech firms send their new models for government assessment 30 days ahead of releasing them. The participation was entirely voluntary. In its final form, the order scrapped previous proposals that would have forced AI models to undergo government approval or licensing prior to their launch.

David Sacks, a venture capitalist who advised the Trump administration on AI, had been part of crafting the policy and still has ties with the president. The move against Anthropic has prompted concerns within the administration about the government being secretly developing an approval process the order sought to avoid.

According to Kush Desai, spokesman for the White House, the administration is working with leaders from the AI industry to ensure that innovation and national security interests are taken into account.

Kush also said, “The United States is by far the world leader in the global AI race, and President Trump is committed to ensuring America’s technological dominance.”

One administration official said the conflict may still fit the executive order if the restriction ends quickly. “If this blows over, if the restrictions are lifted tomorrow, then I would say, ‘Oh OK, this is still kind of consistent with the executive order, and we’ll just keep going for it,’” the official said.

The same person drew a harder line if the ban remains. “If the situation is still f***** a week from now, then I think we have a clear understanding of what’s going on.”

That official allegedly called a prolonged clash “a huge problem” for the U.S. AI sector and warned that every future model developer could end up asking Washington for permission before launch. “That’s an extremely bad situation, and it would completely cripple the whole industry,” the official said. “It would really put the whole AI industry at a disadvantage and be a recruiting and retention problem for AI researchers.”