Anthropic Pulls Fable 5 and Mythos 5 on US National Security Order

Anthropic said on Friday June 12, 2026 it has suspended access to Claude Fable 5 and Claude Mythos 5 for every customer, three days after the public launch of Fable 5. The suspension came in response to an export control directive from the U.S. Commerce Department, delivered by Commerce Secretary Howard Lutnick to Anthropic CEO Dario Amodei.

It is the first time a leading AI company has taken a publicly deployed model offline in response to intervention from the federal government, per NBC News. The directive, which Anthropic published on its own site late Friday, covers any foreign national inside or outside the United States, including foreign national Anthropic employees, and requires a Commerce license for export, re-export, or domestic transfer of either model.

What the Letter From Commerce Says

The letter from Lutnick frames Fable 5 and Mythos 5 as items subject to U.S. export controls, per the Lutnick letter to Amodei. It applies to any location outside the United States and to all foreign persons within the country.

Per the letter, Anthropic will have to submit an additional application for individually validated licenses, and failure to comply would result in financial and civil penalties. Wired reported the directive landed in Anthropic’s inbox at 5:21 p.m. Pacific time on Friday, and the company disabled access to both models within hours.

The directive is narrow in target and wide in effect. Because Fable 5 and Mythos 5 run on a single account system that does not cleanly separate foreign national users, Anthropic decided to cut off all customers rather than try to filter by nationality. “The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance,” the company wrote in its Friday statement.

Why the Government Moved Now

Friday’s letter did not name the specific concern, but an administration official told Axios the Commerce Department acted after a separate company claimed it was able to jailbreak Mythos 5. The official said the resulting “national security risks” prompted the action, and that the model needs to remain locked down until the U.S. government’s national security apparatus is hardened, which the official said could happen in the next few weeks.

That framing places the model in the same export bucket as advanced semiconductors and other items the Commerce Department restricts for foreign buyers. The agency had tried to get Anthropic to pause the release of Fable 5 before launch, the official added, and the export letter followed when Anthropic did not.

Anthropic, in its own statement, said it does not believe the jailbreak gives hackers capabilities that are not already available through other public models. “None of our security testers have found a ‘universal jailbreak,’ or a way to bypass its safeguards against helping hackers,” the company wrote in its Friday statement, as posted on its Fable 5 launch and suspension announcement. NBC News, citing the same statement, reported the company believes the vulnerability “is likely also present in AI systems developed by other companies.”

The two affected models are the same underlying system, marketed under two names with different guardrails:

Anthropic’s Public Disagreement

Anthropic’s response was unusually direct. The company called the action a “misunderstanding” in a post on X and asked the public to weigh in on its principle: a narrow jailbreak, the company said, should not be grounds for recalling a model that is already deployed to hundreds of millions of users.

“We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people,” Anthropic wrote in its Friday statement, as carried by NBC News. “If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers.”

We believe the government should have the ability to block unsafe deployments, as part of a statutory process that is transparent, fair, clear, and grounded in technical facts. This action does not adhere to those principles.

The full statement, which Anthropic published on its own site late Friday, also argued that the directive overreaches because the underlying vulnerability is “likely also present in AI systems developed by other companies,” per Anthropic’s full Friday statement. The administration’s letter, Anthropic noted, did not specify what concrete national security risk the jailbreak presented.

From a Pentagon Blacklist to a Friday Night Letter

Friday’s directive is the latest move in a months-long fight between Anthropic and the Trump administration over how the U.S. government can use the company’s AI. The first major flashpoint came in February 2026, when President Trump ordered every federal agency to stop using Anthropic’s products and Defense Secretary Pete Hegseth followed by labeling Anthropic a supply-chain risk to national security, a designation that bars military contractors and suppliers from working with the company. Trump called the dispute a “DISASTROUS MISTAKE” by “The Leftwing nut jobs at Anthropic.”

Anthropic’s offense, in the company’s own framing, was refusing to drop two restrictions: a bar on using Claude to power fully autonomous weapons, and a bar on using it for mass domestic surveillance of Americans. “We do not believe that today’s frontier AI models are reliable enough to be used in fully autonomous weapons,” CEO Dario Amodei wrote at the time, per the EFF’s analysis of the dispute. “And we believe that mass domestic surveillance of Americans constitutes a violation of fundamental rights.”

The clash is now moving on a separate track. The February blacklisting bars Anthropic from Pentagon work. Friday’s export letter bars foreign access to the company’s best models. The earlier fight sent OpenAI to the Pentagon: hours after Trump’s February directive, OpenAI announced its own classified networks deal with the Defense Department, a window Anthropic had refused to open. The new fight sends Fable 5 and Mythos 5 offline for every non-U.S. Anthropic customer, and for U.S. customers who are foreign nationals.

A federal judge in California sided with Anthropic in a related suit against the February designation, per NBC News, but the case is still active in a Washington, D.C. federal court. In between the two court dates, the company quietly signed on to Project Glasswing, an April partnership that handed a preview of Mythos to a small group of U.S. cyber defenders and critical infrastructure providers.

The Commerce directive now sits on top of that sequence:

1. April 2026: Anthropic begins Project Glasswing, releasing Claude Mythos Preview to a small group of U.S. cyber defenders.
2. June 2, 2026: Trump signs a landmark AI executive order that creates a pre-deployment testing mechanism and a voluntary early-access channel.
3. June 9, 2026: Anthropic publicly launches Claude Fable 5, its first Mythos-class model available to all customers.
4. June 12, 2026: Commerce Secretary Howard Lutnick sends Anthropic an export control letter banning foreign national access to both models.
5. Late June 12, 2026: Anthropic disables Fable 5 and Mythos 5 for all customers and posts a public statement calling the move a “misunderstanding.”

Every Customer, Foreign and Domestic

The directive, by Anthropic’s own reading, covers every customer who is a foreign national, plus any access to either model from outside the United States. The company said late Friday the simplest path to compliance was to disable both models for everyone. “Fable 5 is temporarily unavailable,” the Claude landing page read as of Friday evening.

Those customers include the heavy enterprise users Anthropic featured in its Fable 5 launch post. Stripe reported that Fable 5 compressed months of engineering into days during early testing, including a 50-million-line Ruby codebase migration in a day that would have otherwise taken a whole team more than two months. Cognition, GitHub, Hebbia, IMC, and Box all posted benchmarks or anecdotes backing the new model. Each of those relationships is now on pause.

• 3 days: the gap between Fable 5’s June 9 launch and its June 12 suspension.
• $10 input, $50 output per million tokens: the price Anthropic set for both models, less than half the cost of Claude Mythos Preview.
• Less than 5% of sessions: how often Fable 5’s safety classifiers trigger, per Anthropic’s own launch disclosure.
• $200 million: the value of the Pentagon contract Anthropic lost in February, before this latest move.

Stakes for a $380 Billion IPO

Friday’s directive lands on a company that is in the middle of preparing to go public. Anthropic was valued at $380 billion as of February 2026, when NPR profiled the earlier dispute, and the company has continued to raise its valuation in private markets since then. The Pentagon contract that was lost in February was worth up to $200 million, a fraction of Anthropic’s $14 billion in revenue at the time, but the message it sent to enterprise customers was bigger than the contract value.

Friday’s directive is a different kind of blow. Fable 5 was Anthropic’s first Mythos-class model available to the general public, and Anthropic’s own launch post said its capabilities “exceed those of any model we’ve ever made generally available.” Pulling it three days after launch, from every customer in the world, raises a sharper question: if the U.S. government can disable a frontier model globally with one letter, what does Anthropic’s enterprise roadmap look like in the second half of 2026?

Per an administration official, the model only needs to remain locked down until the U.S. government’s national security apparatus is hardened, which the official said could happen in the next few weeks. Anthropic is calling for a different fix: any future block, the company said, should be grounded in “a statutory process that is transparent, fair, clear, and grounded in technical facts.” The February fight with the Pentagon is the obvious reference point, and it is the fight Anthropic is now fighting again, this time on the Commerce side of the building rather than the Pentagon side. Per the same NPR reporting, the Pentagon fight has not stopped Fable 5’s safety guardrails on launch day or the broader Anthropic enterprise and IPO momentum from advancing, but it does show how much the company is now balancing two regulators at once.

Frequently Asked Questions

Why did Anthropic pull Claude Fable 5 and Claude Mythos 5?

Anthropic said on Friday June 12, 2026 that it had received an export control directive from U.S. Commerce Secretary Howard Lutnick banning any foreign national, including the company’s own foreign national employees, from accessing the two models. To comply, Anthropic disabled Fable 5 and Mythos 5 for all customers, not just foreign nationals.

What triggered the export letter?

Per an administration official who spoke to Axios, the Commerce Department acted after a separate company claimed it had found a method to jailbreak Mythos 5. The official said the directive should remain in place until the U.S. government’s national security apparatus is hardened, which the official said could happen in the next few weeks.

Is this connected to Anthropic’s February fight with the Pentagon?

It sits in the same fight, but on a different track. In February 2026, the Pentagon designated Anthropic a supply-chain risk after Anthropic refused to allow Claude to be used for mass domestic surveillance or fully autonomous weapons. Friday’s directive is a Commerce Department action that covers foreign access to the company’s models, not its federal contracts.

When will Fable 5 and Mythos 5 come back online?

Anthropic has not given a date. The company said it is working to restore access as soon as possible and called the action a misunderstanding. An administration official told Axios the model needs to remain locked down until the U.S. government’s national security apparatus is hardened, which the official said could happen in the next few weeks.