The Trump administration on Friday partially rescinded its export ban on Anthropic’s most advanced artificial intelligence model — deescalating a confrontation that has caused confusion across the American AI industry.
The release clears the way for a select group of more than 100 companies and agencies to gain access to the Mythos 5 model, two weeks after the administration imposed restrictions amid fears that the software could be used to launch cyberattacks. But a second advanced Anthropic model, called Fable 5, remains blocked.
Meanwhile, pressure from the White House led Anthropic’s leading competitor, OpenAI, to limit the release of its most advanced model this week because of similar cyber concerns.
The series of back-and-forth actions on AI, from an administration that has promised to unleash American companies to dominate the technology, has caused concerns in the industry about how heavy a hand President Donald Trump plans to take in overseeing what kinds of products U.S. developers can release to the market. It’s been less than four weeks since Trump signed an executive order rejecting mandatory federal controls on AI, instead asking leading tech companies to submit new models to a voluntary government review before making them widely available.
Conversations between Anthropic and the administration are expected to continue through the weekend on restoring access to Fable, according to a person who was granted anonymity to describe the private discussions.
The conversations have also focused on determining a standardized framework to evaluate cases of suspected security bypasses in the future, the person said, beyond just addressing the recent episode. POLITICO reported last week on discussions of a technical assessment that would create ground rules for these kinds of disputes.
“Since the issuance of my June 12 letter, Anthropic has worked with the U.S. government to address risks associated with the Covered Models,” Commerce Secretary Howard Lutnick wrote in a Friday letter to Anthropic’s chief compute officer Tom Brown, obtained by POLITICO. “These efforts have yielded significant progress.”
Semafor first reported on the new letter Friday night.
Anthropic, which has disputed the concerns about its cyber controls, issued a statement later Friday welcoming the administration’s turnabout.
“We received notice from the U.S. government that Mythos 5, our strongest cybersecurity model, can be redeployed to a small group of cyber defenders and infrastructure providers,” the company said. “We are working to provision the approved set of providers and restore their access to Mythos 5 as quickly as possible. We are pleased to see this progress and continue to work with the government to expand access to Mythos 5 and make Fable 5 available for general use again.”
But others across the industry have expressed a desire for more clarity about the administration’s long-term intentions. Those included OpenAI, which announced Friday that, at the administration’s request, it is making only a limited release of three versions of its new GPT‑5.6 model.
“We don’t believe this kind of government access process should become the long-term default,” OpenAI said in a blog post Friday. “It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them.”
Friday’s reversal on Mythos came after Anthropic staff and senior White House officials met over the last two weeks to discuss ways to defuse the dispute, including technical questions about whether the company had done enough to prevent its AI products from being used to launch cyberattacks. The action allows the company to restore access to its Mythos 5 models to a list of more than 100 “trusted partners,” including companies and federal agencies approved by the government, an administration official familiar with the move granted anonymity to discuss it said.
“In just two weeks, we have worked diligently to ensure America remains the global leader in AI while safeguarding our security,” said Benno Kass, a Commerce Department spokesperson.
The administration had slapped the export restrictions on Fable and Mythos, banning their use by foreign nationals, after a series of tense calls with Anthropic CEO Dario Amodei about whether the company’s safety guardrails were as solid as promised. Anthropic then announced it was cutting off access “for all our customers to ensure compliance” — while insisting that “we believe this is a misunderstanding.”
Amodei and Trump later sat across from each other during a G7 lunch with allied leaders and AI executives last week in Évian-les-Bains, France.
The administration’s frantic effort to address the safety issue raised a host of questions about how it would deal with future security concerns in the rapidly evolving AI industry, as well as whether Trump’s aides were unfairly singling out Anthropic after previous clashes with the company. It also left U.S. allies in Europe and Canada abruptly unable to use the new Anthropic models to identify and patch their own cybersecurity vulnerabilities, renewing calls for the EU to lessen its reliance on American technology.
Meanwhile, reports of rapid advances in Chinese AI technology have raised worries across the industry about whether U.S. companies will lose their competitive advantage while waiting for administration approvals.
