At 5:21 p.m. ET on June 12, Anthropic received a letter from the US government that the company says forced it to do something no major AI lab has done before. Within hours, Claude’s landing page showed that one of its newest and most capable models was simply unavailable.
The order came from the Commerce Department, in a letter signed by Secretary Howard Lutnick, and it cited national security authorities to demand that Anthropic suspend all access to two of its models, Fable 5 and Mythos 5, for any foreign national, whether inside or outside the United States, including Anthropic’s own foreign-national employees, according to Anthropic’s official statement.
The timing makes the directive particularly striking. Anthropic had announced Fable 5 and Mythos 5 only days earlier, touting both as state of the art across a number of industry benchmarks.
Fable 5 is built on top of Mythos, the underlying model designed with especially strong cybersecurity capabilities, including the ability to identify vulnerabilities in code and explore ways to exploit them, capabilities Anthropic has marketed as valuable for defensive security work.
Why Anthropic disabled the models for everyone, not just foreign users
The directive technically targeted foreign nationals, but Anthropic said it has no way to reliably separate those users from everyone else in real time. The practical result, the company said, was a hard shutoff of both models worldwide.
“The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance,” Anthropic said. Access to all of Anthropic’s other models, including Claude’s standard lineup, was not affected.
More AI:
- Micron sits at the center of a red-hot chip rally
- IBM CEO sends blunt message on AI and quantum computing
- Anthropic CEO makes shocking admission about AI
The move marks the first time the Trump administration has used an export control directive to force a leading AI company to pull a publicly deployed model from its own customers, rather than restrict future sales or exports, according to Bloomberg,
Export control directives typically govern future transactions, blocking a company from selling specific technology to specific buyers going forward.
Applying that framework retroactively to a model already running in production, used by ordinary paying customers around the world, is what made Anthropic’s compliance look less like a sales restriction and more like a product recall.
The jailbreak claim at the center of the dispute
Anthropic said the government’s letter did not provide specific details of its national security concern, but the company’s understanding is that officials became aware of a method of bypassing, or “jailbreaking,” Fable 5.
According to CNBC, an administration official told Axios that the Commerce Department acted after another company claimed it had successfully jailbroken Mythos.
Anthropic disputed the severity of that claim directly.
The company said it reviewed a report it believes is the basis for the directive and found the technique amounted to asking the model to read a specific codebase and identify software flaws, a capability the company says is “widely available from other models, including OpenAI’s GPT-5.5, and is used every day by the defenders who keep systems safe.”
“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 said in its statement.
This is not the first time the administration tried to stop these models
The export control directive was not the administration’s first attempt to intervene with Fable 5 and Mythos 5. Axios reported, via 9to5Mac, that the Trump administration had previously tried to stop Anthropic from releasing the models in the first place, but did not succeed at the time.
Both models trace back to Claude Mythos Preview, an advanced system Anthropic initially made available only to a small group of companies and research partners through a program called Project Glasswing, where participants used it to identify and fix security vulnerabilities. Fable 5 and Mythos 5 were the broader commercial releases that followed.
The relationship between Anthropic and the administration had appeared to be improving in the weeks before this directive. Earlier this year, a high-profile clash with the Department of Defense played out publicly before the two sides reached a negotiated resolution.
On June 2, just ten days before this latest directive, Trump signed an executive order directing federal agencies to strengthen government cyber defenses and establish a mechanism for voluntary early access to the most powerful models from companies including Anthropic and OpenAI.
Yeh/Getty Images
What Anthropic is asking for instead
Anthropic’s response did not reject the principle of government oversight of frontier AI models. The company pointed to its own previously stated position that the government should be able to block unsafe deployments, but only as part of a statutory process that is transparent, fair, clear, and grounded in technical facts.
“This action does not adhere to those principles,” Anthropic said, while confirming it was complying with the directive, calling the situation a likely misunderstanding, and saying it was working to restore access as soon as possible.
Other angles developing around this story
- AI critic Gary Marcus called the move contradictory, arguing that restricting Chinese-born researchers’ access to frontier models could push some of them to return to China, working against the administration’s own stated goal of staying ahead of China in AI, according to Fortune.
- Just last week, the Financial Times reported that the NSA was already using Mythos to conduct offensive cyberattacks, a detail that adds friction to the government’s framing of the model as primarily a security risk, according to NBC News.
- Anthropic is reportedly preparing for a public listing, having hired law firm Wilson Sonsini for IPO work that could begin as early as 2026, a process that regulatory disputes like this one could complicate.
- Anthropic is currently suing the Trump administration over an earlier Pentagon blacklisting dispute, and that litigation is still ongoing separately from this export control directive.
What this means for the AI industry’s relationship with Washington
For an industry that has spent the past two years lobbying for a predictable regulatory framework, this episode is a test case for what government intervention actually looks like in practice when it arrives outside that framework.
An export control directive, typically associated with restricting sales of advanced chips or software to foreign adversaries, was instead used to force the removal of a model already in the hands of paying customers.
Whether this becomes a template for future action against other frontier AI labs, or an isolated dispute that gets resolved within days, will depend largely on what happens next between Anthropic and the Commerce Department.
For now, two of the most capable AI models commercially available went dark for every Anthropic customer within hours of a single letter, regardless of where those customers are located or what they were using the models for.
Related: Anthropic just landed one of the biggest deals in AI
