In Evian-les-Bains, France, there’s news that the Trump administration is open to negotiations with Anthropic regarding a recent order that restricts foreign access to its advanced AI model, according to a report. A high-ranking official in the White House, preferring to remain unnamed, hinted that the company might be able to engage directly with CEO Dario Amodei to find a resolution.
“We can’t let the Frontier Model run rampant,” a source informed on the situation said.
It seems that these comments suggest the government could be approaching last week’s export control directive as a starting point for discussions, rather than an enduring prohibition. This development surfaces amid the fallout from Anthropic’s announcement that they must deactivate their Fable 5 and Mythos 5 models globally due to national security concerns raised by the Commerce Department.
The directive required Anthropic to limit access to users in the United States, which the company found unfeasible, leading to their decision to disable the models worldwide.
Additionally, the U.S. government turned down a request from British Minister Keir Starmer for exemptions for British nationals and companies, as reported on Monday.
Anthropic has also indicated that officials have mentioned a possible “jailbreak” scenario where the model could unintentionally expose software vulnerabilities, though the company remarked that they received only verbal details on this limited aspect.
Anthropic firmly rejected the notion that discovering a potential jailbreak should lead to the withdrawal of a model utilized by millions. “We do not agree that the discovery of a narrow jailbreak possibility should be the cause for a recall of a commercial model that has been deployed to hundreds of millions of people,” they stated after the restrictions were enforced.
The agency later suggested that the confusion around the directive might have been a misunderstanding, and they are currently working towards restoring access.
This closure comes shortly after the launch of Anthropic’s latest model, Fable 5, indicating some significant tension. The company was previously collaborating with the government on safety protocols in anticipation of this release.
Strained relations between Anthropic and the Trump administration have been ongoing for several months. Earlier in the year, the administration placed Anthropic on a supply chain blacklist when the company declined to let the U.S. military use its models for surveillance or autonomous weaponry.
An AI expert from Anthropic expressed concern that if similar restrictions were implemented across the industry, it would impede the progress of new frontier models from all major AI developers. Meanwhile, other Anthropic models remain accessible.
The newspaper has reached out to the White House press secretary for further comment.
