Anthropic PBC won US approval to restore some access to its powerful Mythos 5 artificial intelligence model, after resolving Trump administration concerns about the technology’s potential threats to national security.
“Anthropic has worked with the US government to address risks associated with the Covered Models,” Commerce Secretary Howard Lutnick wrote to the company’s chief compute officer in a letter seen by Bloomberg News. “These efforts have yielded significant progress,” he wrote, saying that the model could be released to “certain trusted partners.”
The clearance eases a confrontation that erupted two weeks ago when the government abruptly barred Anthropic from giving foreign nationals access to Mythos 5 and a related model, Fable 5, over fears that security guardrails could be circumvented. The order prompted the company to disable all global access to the two advanced systems and touched off intensive talks toward a resolution.
“In just two weeks, we have worked diligently to ensure America remains the global leader in AI while safeguarding our security,” said Commerce Department spokesman Benno Kass.
The Commerce letter dated Friday doesn’t mention any change to its restrictions on use of the Fable 5 model.
“We received notice from the US government that Mythos 5, our strongest cybersecurity model, can be redeployed to a small group of cyber defenders and infrastructure providers,” Anthropic said in a statement. “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.”
It’s unclear what steps Anthropic had to take to satisfy the government’s concerns that it was possible to “jailbreak,” or bypass guardrails on, the advanced models. The company expects conversations to continue over the weekend with the government as it aims to bring back access to Fable 5 soon, said a person familiar with the matter, who asked not to be named because the discussions are private.
The company is also working with the government to come up with a policy framework to deal with such situations in the future, the person said.
Anthropic had objected to the government’s decision to impose export controls in a blog post announcing that the two systems had been disabled. “If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers.”
In that post, Anthropic cautioned that it believes that “perfect jailbreak resistance is not currently possible for any model provider.”
The government’s decision to restrict the availability of Anthropic’s models also reverberated across the AI industry, particularly at rival startup OpenAI, which limited the release of a capable new AI model — GPT-5.6 — under pressure from the Trump administration. OpenAI said Friday it would roll out a preview version to select partners who were approved by the US government, before making it more available in the coming weeks, following a request from the administration.
Wider feud
The showdown over Mythos 5 and Fable 5 marked the latest clash between Anthropic and the Trump administration. Anthropic has been embroiled for months in a feud with the Pentagon over extra guardrails the San Francisco-based company sought for military use of its AI tools. After contract talks broke down, the Defense Department declared the firm a supply-chain risk in March and is moving to seek other AI providers for the armed forces.
Privately held Anthropic, which has long positioned itself as a more responsible AI developer than its rivals, first shared its Mythos model in April with a limited group of companies and institutions, warning that its ability to find cybersecurity vulnerabilities made it too risky to distribute more widely.
Earlier this month, the company released Fable 5 as the first public-facing version of its Mythos-class model, but with guardrails aimed at containing the full range of its cyber capabilities. Fable 5 is prevented from responding to certain types of queries, including those related to cybersecurity and biology. In those cases, Anthropic has said its Claude chatbot will route responses through a different model, called Opus 4.8.
By continuing to restrict the Fable 5 model, the federal government indicated that it remains wary of making the technology behind these models broadly available, even with limits that Anthropic added before generally releasing it.
Mythos 5, which is a version of Fable 5 without some of the safeguards, was scheduled to be released to a more limited group of institutions for cybersecurity purposes, an invite-only initiative called Project Glasswing. As of earlier this month, about 200 firms had access to Mythos through the program, including Apple Inc., Alphabet Inc.’s Google, Cisco Systems Inc., Nvidia Corp. and Microsoft Corp.
Industry impact
The order to bar Anthropic’s latest models represented the most significant intervention by the US government to date into an AI venture’s operations, and highlighted emerging concerns surrounding the security of cutting-edge models. It posed a new challenge to Anthropic weeks after the company filed confidentially for an initial public offering, with its latest valuation topping $900 billion.
Read More: Anthropic Files Confidentially for IPO in Race With OpenAI
The recent negotiations involved senior executives from Anthropic including co-founder Tom Brown, who met with Lutnick and other senior administration officials in recent days, people familiar with the talks told Bloomberg. US officials told Anthropic that the restrictions would be lifted once the government’s security concerns have been addressed, said the people, who spoke on condition of anonymity because the discussions weren’t public.
In the talks with Commerce over Fable 5 and Mythos 5, Anthropic Chief Executive Officer Dario Amodei took a hands-off role, helping to reduce friction between the two sides, according to the people.
This story was originally featured on Fortune.com