Anthropic Wins Partial Mythos 5 Clearance as Fable 5 Stays Locked

Anthropic’s Claude Mythos 5 cleared its biggest U.S. regulatory hurdle on Friday, with the Commerce Department greenlighting a limited release to roughly 100 trusted companies and federal agencies. The letter from Commerce Secretary Howard Lutnick, addressed to Anthropic co-founder Tom Brown, leaves the consumer-facing Fable 5 model frozen and the broader standoff with the Trump administration unresolved.

Two weeks of export-control brinkmanship bought Anthropic a partial victory and a louder fight over Fable 5. OpenAI, by contrast, ran a quieter rollout of GPT-5.6 Sol the same day and used a different playbook that pre-coordinated with Washington before launch.

What Lutnick’s Letter Permits

The June 26 letter lets Anthropic share Claude Mythos 5 with the entities listed in an attached “Annex A” and with their foreign national employees. It also lifts the same restriction on Anthropic’s own foreign national staff. The export-control directive from June 12 stays in force for everyone else.

I have determined that appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model.

Commerce Secretary Howard Lutnick wrote those words in a letter addressed to Tom Brown, Anthropic’s co-founder and chief compute officer, the same executive who has taken over the company’s White House talks. Lutnick framed the move as a calibrated trust decision rather than a full release, citing “significant progress” in the company’s engagement with the Commerce Department. The letter does not name who is on Annex A, but the count of roughly 100 institutions lines up with the size of Anthropic’s pre-existing Project Glasswing program for cyber defenders and critical infrastructure operators. Anthropic declined to make Brown available for an interview, but the company pointed to the letter as the opening wedge it had been working toward since the June 12 directive. The Lutnick letter, like the directive, applies only to Mythos 5; Fable 5 is not addressed.

Anthropic spokesperson Eduardo Maia Silva said the company “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,” and that the company is “working to provision the approved set of providers and restore their access to Mythos 5 as quickly as possible.” Silva added that Anthropic is “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.” That second sentence is the closest the company has come to a public timeline on Fable 5.

• Letter date: June 26, 2026
• Signer: Commerce Secretary Howard Lutnick
• Addressee: Anthropic co-founder and chief compute officer Tom Brown
• Model cleared: Claude Mythos 5
• Model not addressed: Claude Fable 5
• Recipients: Roughly 100 US companies and federal agencies

Why Fable 5 Stays Out of Reach

The Lutnick letter is silent on Fable 5, the version of Mythos that Anthropic released to the public on June 9 with three layers of guardrails. That missing sentence is the reason consumer access remains blocked, even after the more restricted cyber model got a green light. The same silence is why Fable 5 is not on the Annex A list, even though some of its users are the same federal agencies now cleared for Mythos 5. Anthropic has not named a date for restoring Fable 5 access, but the company told WIRED it is still in talks with the Commerce Department about the consumer model. Claude Fable 5’s public release on June 9 marked the first time Anthropic shipped a model that capable to consumers at all.

Anthropic said in a statement to WIRED that it is still working to “expand access to Mythos 5 and make Fable 5 available for general use again.” People close to the talks told Semafor that the two sides are moving toward releasing Fable as well, though there is no firm timeline on the table. The two Anthropic models split along cyber-capability lines. Fable 5, the consumer release, blocked responses in specific high-risk areas. Mythos 5, the more capable cousin, was always limited to vetted cyber defenders through Project Glasswing. The export-control directive froze both, but the Lutnick letter restores only the more restricted of the two, a sequencing that tells users the consumer release is the harder lift, not the easier one.

The Two-Week Export-Control Fight

On June 12, the Commerce Department sent Anthropic an export-control directive that ordered the company to suspend access to Fable 5 and Mythos 5 by “any foreign national, whether in or outside the United States, including foreign national Anthropic employees.” Anthropic said the order forced it to disable both models for every customer, and it did so within hours. The Trump AI executive order that built a voluntary cyber framework had already set the policy backdrop, and the directive was the first time those rules hit a live frontier model.

The directive is, on its face, the first time US export controls have been used to control access to a deployed AI model in this way. “To my knowledge, this is the first time US export controls have been used to control access to an AI model in this way,” Hanna Dohmen, a senior research analyst at Georgetown University’s Center for Security and Emerging Technology, told The Verge. The order did not include a public legal justification, and Anthropic noted that the order also affected its own foreign national researchers. Read alongside the original disable of Fable 5 and Mythos 5, the directive set the stage for everything that followed.

Behind the order sat a chain of warnings. WIRED reported that the Trump administration grew concerned after learning Anthropic had granted Mythos access to a South Korean telecommunications firm that officials believed had ties to China. Amazon and the National Security Agency separately raised concerns to the White House that Fable 5 could be jailbroken, and the confluence of those events convinced officials they had to act.

Anthropic pushed back on the technical premise. The company said the alleged jailbreak amounted to asking a model to read a specific codebase and fix software flaws, and that the capability it showed was “widely available from other models (including OpenAI’s GPT-5.5), and is used every day by the defenders who keep systems safe.” Brad Carson, president of Americans for Responsible Innovation, said the episode showed AI regulation must run “consistently across industry, without favor, and according to a clear, rules-based process.” The first-time use of US export controls on an AI model left the rest of the industry guessing whether it would be next.

• June 9, 2026: Anthropic releases Fable 5 to the public and Mythos 5 to Project Glasswing partners.
• June 12, 2026: Commerce Department issues the export-control directive.
• June 13, 2026: Anthropic disables both models for all customers.
• June 26, 2026: Lutnick letter clears Mythos 5 for the Annex A partner list.

Tom Brown Replaces Dario Amodei in the Talks

The Lutnick letter is addressed to Tom Brown, not to CEO Dario Amodei. That detail is not incidental. WIRED reported that the White House, frustrated with talks, had described Amodei as a “weirdo” in private, and that Brown had taken over the company’s high-stakes meetings with the administration. An anonymous person involved in the calls told WIRED: “Tom Brown is not being a weirdo like Dario and can actually engage.” The personnel change at the center of the talks is the single most cited reason the two sides moved from confrontation to a partial clearance inside two weeks.

Brown has not been alone in Washington. WIRED reported that Anthropic sent senior members of its cybersecurity and AI safety teams to the capital, and that public policy chief Sarah Heck has been leading the company’s discussions with the Commerce Department alongside Brown. Anthropic declined to confirm the personnel change on the record, but the choice of addressee on the Lutnick letter reflects who the government is now talking to, and who it is not. The same week, White House officials also told WIRED that the cleared companies and federal agencies on Annex A may now allow their own foreign national employees to access Mythos 5, a concession that effectively revives Anthropic’s full research workforce inside the United States.

OpenAI’s Coordinated Rollout

On the same Friday, OpenAI launched GPT-5.6 Sol, Terra, and Luna to a small group of trusted partners whose participation, the company said, had been shared with the government. Sol is OpenAI’s flagship, Terra a balanced model for everyday work, and Luna a fast and affordable option. The company previewed the models’ capabilities to Washington before launch, and it plans to make all three generally available “in the coming weeks.” The same-day preview of GPT-5.6 Sol, Terra, and Luna ran in parallel with the Lutnick letter, a coincidence that made the contrast between the two labs hard to miss.

OpenAI was explicit about not wanting this to become routine. The company wrote that the arrangement should not become the long-term default, and it framed the cooperation as a short-term step designed to clear the path to broader release.

We don’t believe this kind of government access process should become the long-term default. It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them.

OpenAI published that preview on its own site, signed by the OpenAI team, and the company said it was taking the short-term step because it believed the cooperation was the strongest path to broader availability in the coming weeks. The strategic contrast with Anthropic is sharp. Anthropic released its strongest cyber model without pre-coordination, got hit with an export-control directive, and is now negotiating its way back one model at a time. OpenAI pre-coordinated, launched under the same restricted-trusted-partner frame, and emerged with a smoother public-relations footprint and a public statement of intent to broaden access. The Trump administration is, in effect, rewarding the lab that talked to it first and penalizing the one that did not.

The Pentagon Blacklist Case Rolls On

The Mythos dispute is the latest round in a longer confrontation between Anthropic and the federal government. Earlier in 2026, the Department of Defense clashed with the company over its refusal to allow Claude to be used for fully autonomous weapons and mass domestic surveillance. The Pentagon responded by designating Anthropic a supply chain risk to U.S. national security, a label that has historically been reserved for foreign adversaries and that forces defense contractors to certify they will not use the company’s models in their work with the military.

Anthropic sued the Trump administration to overturn the designation, and a federal judge issued a temporary injunction on March 27. An appeals court denied Anthropic’s motion to lift the designation on April 8, and the litigation is still pending. The Lutnick letter does not touch the supply-chain case, and Anthropic’s earlier statement to the Department of War still stands: the company says it will provide its models to the Department and the national security community “at nominal cost and with continuing support from our engineers, for as long as is necessary to make that transition, and for as long as we are permitted to do so.” The supply-chain-risk question now sits alongside the unresolved Fable 5 freeze as two open fronts. Fable 5 will need a separate Commerce Department letter to clear the same gate Mythos 5 just walked through, and the DoD case will need a court ruling, not a letter, to settle the longer fight over Anthropic’s place in the federal stack.

Frequently Asked Questions

What is Claude Mythos 5?

Claude Mythos 5 is Anthropic’s most capable cybersecurity-focused AI model, designed to help defenders find and fix software flaws in critical systems. It was released in early June 2026 to a vetted group of cyber defenders and critical infrastructure operators through a program called Project Glasswing. Anthropic has called it the strongest cybersecurity model it has ever shipped, and the Lutnick letter is the first formal US clearance for it.

Why was Claude Mythos 5 suspended?

The Commerce Department sent Anthropic an export-control directive on June 12, 2026, that required the company to cut off access to Mythos 5 and Fable 5 for any foreign national, including foreign national employees inside the United States. The directive cited national security authorities and followed separate warnings from Amazon and the National Security Agency about a potential jailbreak of Fable 5, plus concerns about Mythos access at a South Korean telecom firm with alleged ties to China.

Is Claude Fable 5 still available?

No. The Lutnick letter of June 26 is silent on Fable 5, and the consumer model remains suspended for all customers. Anthropic has said it is still in discussions with the White House about restoring Fable 5 access, and people close to the talks told Semafor the two sides are moving toward a release, though no timeline is public.

Who can now use Claude Mythos 5?

Under the Lutnick letter, Mythos 5 can be exported and re-exported to the entities listed in Annex A of the letter, to their foreign national employees, and to Anthropic’s own foreign national employees. Roughly 100 US companies and federal agencies are covered, drawn largely from the existing Project Glasswing coalition, and the list is titled “Anthropic US Entities – Approved.”

When will Fable 5 come back online?

Anthropic has not named a date. The company says it continues to work with the US government to “expand access to Mythos 5 and make Fable 5 available for general use again,” and talks were expected to continue through the weekend of June 27. Until the Commerce Department issues a separate letter for Fable 5, the consumer model remains frozen.