Trump officials meet with Anthropic to discuss a truce

• Anthropic met with Trump officials to resolve the export ban on Fable, POLITICO reported.
• Anthropic argued the breach was minor, not a "jailbreak."
• AI experts warn export controls may hinder US AI leadership, impacting innovation and competition.

Anthropic staff met Monday with senior Trump administration officials for their first in-person sit-downs after a federally imposed export ban forced the artificial intelligence startup to pull its latest model from the market Friday night, two administration officials, a person familiar with the discussions and a person close to the company told POLITICO, which is — like Business Insider — part of the Axel Springer Global Reporters Network.

It will likely take longer than a few days to reach a resolution that eases the federal government's Friday action, which had barred Anthropic from allowing non-U.S. users to access its newest model because of potential security vulnerabilities, a senior White House official said. But the official left the door open to the possibility that it can be done quickly.

"That's up to Anthropic," the official said.

The company raced to send senior leaders with research and safeguard expertise to D.C. after multiple hourslong calls among Anthropic co-founder Tom Brown, Commerce Secretary Howard Lutnick and National Cyber Director Sean Cairncross over the weekend, said the officials, who like others quoted in this story were granted anonymity to describe the discussions. Anthropic head of public policy Sarah Heck was also present on those calls.

Monday's in-person meetings, held by Commerce and Cairncross' office, were more technical and were led by staff, including Chris Fall, who heads Commerce's Center for AI Standards and Innovation, the person familiar with the discussions said.

Anthropic gave a presentation to administration officials, explaining Anthropic's cybersecurity safeguards in hopes of moving past the restrictions, the administration official said.

Representatives from Anthropic included Logan Graham, who evaluates and stress-tests models as part of the company's Frontier Red Team; Dave Orr, the company's head of safeguards; and Nicholas Carlini, its lead security researcher, a person close to Anthropic said.

The export restriction was the most significant escalation yet in the administration's attempts to regulate the AI industry, which is developing faster than the government's infrastructure around the groundbreaking technology. It was also the first time the White House has forced a company to remove a model from public access — but only the latest in a series of spats between the administration and Anthropic over ethical restrictions on potentially catastrophic AI risks.

Anthropic has defended its safeguards since the restrictions were handed down Friday night.

"These vulnerabilities all appear relatively simple, and we have found that other publicly-available models are able to discover them as well without requiring a bypass," it said in a blog post.

Treasury Secretary Scott Bessent, who played a leading role in the discussions with Anthropic CEO Dario Amodei before the administration implemented the export restrictions, was less involved over the weekend, and Treasury did not participate in the in-person meetings Monday.

Lutnick was tapped in by the president and became more heavily involved over the weekend because of his oversight of the nation's export policy, the person familiar with the discussions said. Lutnick and Bessent both traveled on Air Force One with the president to France for G7 meetings taking place Monday through Wednesday.

The export control, which banned Anthropic from allowing foreign nationals to use its latest model, Fable 5, effectively forced the company to pull the model down entirely in order to comply, the company said Friday night.

The escalation came after a disagreement between Anthropic and the administration over a perceived security concern with Fable that Amazon, a key investor in Anthropic, had raised to the White House late last week, as POLITICO previously reported.

Amazon reported that it found a way to bypass Anthropic's guardrails on Fable, a long-awaited model designed to be less powerful than the company's highly advanced model Mythos. Anthropic has allowed only a select group of companies to gain access to Mythos because of its ability to exploit cyber vulnerabilities faster than any human.

Fable performed at a level similar to Mythos, but came with built-in safeguards to prevent users from prompting it to carry out a cyberattack or to detail how to create a bioweapon, the company said. Amazon's discovery, elevated to the White House and the National Security Agency, prompted Cairncross, Bessent and others to demand that Anthropic halt access to Fable.

When Anthropic did not agree, arguing instead that the security breach was minor and not a larger so-called "jailbreak" that would allow people to use it for cyberattacks, the administration handed down an export control that the company said forced its hand.

Over the weekend, a group of nearly 80 technical experts and CEOs wrote an open letter to the administration urging officials to revoke the export control. They said the actions that Amazon researchers had gotten Fable to carry out were a "necessary capability in any model that is intended to write secure code," and "should not be considered an offensive capability."

The Trump administration's abrupt controls have "taken the best models away from defenders, created market uncertainty, and risked America's AI leadership without any real risk to justify it," the tech leaders wrote in the letter.

Graham, one of the Anthropic employees attending meetings with the administration, had told Fox News in April that Anthropic was looking to walk a fine line in giving defenders access to Mythos "without the bad actors having access."

"Ultimately, we think that tools like this need to be scaled to the entire world in a secure way because we need the entire world to use these tools to defend themselves as well," Graham added. The interview occurred before Fable's release.

Some in the tech community have vouched for the tough export restrictions and said the risks posed by the guardrails being broken were serious. But other AI experts, and even officials within the administration, have said the controls create a threatening regulatory landscape that will hobble the industry.

"AI is licensed now, but the requirements change constantly and are always a secret, even to the administration itself, which will discover the rules spontaneously in real time as it reacts to events," Dean Ball, a former Trump administration official who worked on AI policy, wrote on X.

One current administration official warned that the longer the export dispute goes on, the more likely it is to become a de facto licensing regime that AI companies would have to contend with when they release new models, slowing their innovation and competition with China.

If the export control doesn't amount to a brief "slap on the wrist" that's quickly alleviated, "it's going to be a huge problem for the entire industry," the administration official said.

"It means that every model going forward needs to ask the government's permission for whether it can be released. That's an extremely bad situation," the official said.

Megan Messerly contributed to this report.

This story originally appeared on POLITICO and is courtesy of the Axel Springer Global Reporters Network, which harnesses the resources of the company's newsrooms to publish ambitious scoops, investigations, interviews, opinion pieces, and analysis. It allows journalists — including those from POLITICO, Business Insider, WELT, BILD, Onet, and Fakt — to collaborate on major stories for an international audience of hundreds of millions across platforms.