The Anthropic Fable 5 saga took a stunning turn just days after the company unveiled its most capable AI models to the public. On the evening of June 12, 2026, only three days into the launch, the U.S. government ordered the models taken offline, setting off a clash that has rippled across the AI industry and raised pointed questions about how much control governments can exert over frontier technology.

The Order That Shut Everything Down

According to reporting from Axios, Commerce Secretary Howard Lutnick sent a letter to Anthropic CEO Dario Amodei directing the company to suspend all access to Claude Fable 5 and Claude Mythos 5 by any foreign national, anywhere in the world. The directive’s reach was sweeping, extending even to foreign nationals employed inside the United States, including Anthropic’s own staff.

Faced with a restriction that broad, the company concluded it had no practical choice but to switch off the models entirely. Within hours, both were offline for every customer worldwide.

In its official statement, Anthropic explained that the export control directive, issued under national security authorities, forced an abrupt shutdown to ensure compliance. Access to all other Claude models, including the latest Claude Opus 4.8, remained unaffected.

What Fable 5 and Mythos 5 Actually Were

For users who noticed two unfamiliar model names appear over the past week, the distinction matters. Fable 5 was the headline launch, representing Anthropic’s first general-release version of its Mythos model family, a lineup the company had previously described as too powerful in cybersecurity to release openly.

The two models share the same underlying architecture but differ in their guardrails:

• Fable 5 included classifiers designed to block responses in high-risk areas such as cybersecurity, making it suitable for public release.
• Mythos 5 operated with some of those constraints removed and was available only to a vetted group of organizations through Anthropic’s Project Glasswing program.

Both, as of the order, are now gone.

The ‘Jailbreak’ at the Center

The trigger appears to have been a reported technique for bypassing Fable 5’s safety guardrails, the filters meant to prevent the model from tapping the advanced cybersecurity capabilities baked into the Mythos architecture. The worry was straightforward: if those guardrails could be defeated, a consumer-facing AI product might effectively become an unrestricted cyber tool.

Axios reported that the administration had earlier tried to stop Anthropic from releasing the models at all and failed, making the export control order the government’s next move.

A day after the shutdown, White House AI adviser David Sacks offered the administration’s account on X. He wrote that a highly credible partner trusted by both Anthropic and the government had come forward with a jailbreak of the guardrails, and that the administration asked Amodei to either fix it or de-deploy the model, only for Amodei to refuse. Sacks characterized the response as “reluctant” and said the resolution was simple, placing the ball squarely in Anthropic’s court.

Anthropic’s Pushback

Anthropic disputes both the severity of the jailbreak and the characterization of its conduct. The company said it received the directive at 5:21 p.m. ET, with the letter providing no specific details of the underlying national security concern.

After reviewing a demonstration of the technique, Anthropic concluded that it surfaced only a small number of previously known, minor vulnerabilities, ones that other publicly deployed models, including OpenAI’s GPT-5.5, can also discover and that cybersecurity defenders use routinely. The company argued that recalling a model deployed to hundreds of millions of people over a narrow potential jailbreak set a dangerous standard, warning that applying it across the industry would essentially halt all new model deployments by frontier providers.

Anthropic also emphasized that before launch, the models underwent thousands of hours of red-teaming by the U.S. government, the U.K. AI Security Institute, and third-party organizations, none of which found a universal jailbreak. A source close to the company contended that Anthropic was never presented with details and never refused to fix anything. The competing accounts remain unreconciled.

Amazon’s Apparent Role

One of the more striking threads concerns who flagged the jailbreak. While the identity of the “trusted partner” has not been officially confirmed, reporting by Semafor and the Wall Street Journal points to Amazon.

CEO Andy Jassy reportedly told Treasury Secretary Scott Bessent and other officials that Amazon researchers used Fable 5 to obtain information that could be used in cyberattacks, after which the government imposed the ban. The detail is notable because Amazon is one of Anthropic’s largest investors and supplies much of its cloud infrastructure, making its purported role in the shutdown especially unexpected. An Amazon spokesperson told Semafor that it isn’t uncommon for governments to seek the company’s counsel on security risks, while declining to share specifics.

Another reported angle suggests the White House acted partly out of concern that a China-linked group had accessed Mythos, raising fears the model could be reverse-engineered or distilled by a foreign adversary.

A Tense History

The shutdown did not emerge from a vacuum. Since early 2025, Anthropic and the Trump administration have been locked in a thorny conflict. The Department of Defense previously designated Anthropic a supply chain risk after talks collapsed, a label historically reserved for foreign adversaries, and Anthropic sued to challenge that designation in a case still active in court.

Defense Secretary Pete Hegseth weighed in on X, claiming the Department had removed Anthropic from its building “forever” and that each passing day proved the decision right. Sacks, for his part, pushed back on the idea that old grievances were driving the export order, insisting that those tying the action to prior disputes were wrong.

The Timing and the Stakes

Several timing details sharpen the story’s significance. Just one day after launching Fable 5, Amodei published a major policy essay urging the U.S. government to hold legal authority to block or reverse frontier AI models that fail independent safety testing. Two days later, the government used precisely that kind of authority against his own company.

The shutdown also arrives at a commercially delicate moment. Anthropic recently filed a confidential IPO prospectus with the SEC, disclosing a revenue run rate of $47 billion and a valuation of $965 billion. The episode could give investors pause, prompting questions about whether the company can stay at the cutting edge if the government continues singling out its models. Anthropic, for its part, said it believes the situation is a misunderstanding and is working to restore access as soon as possible.

A Precedent Worth Watching

The deeper implication is hard to miss. A sitting government has now demonstrated that it can switch off a widely used AI product mid-deployment, based on a security assessment the company says is inaccurate. How that precedent unfolds is something the entire industry will be tracking closely.

Fittingly, Amodei opened his pre-shutdown essay with a metaphor about mismatched speeds, likening the relationship between AI and political institutions to the Hobbits trying to rouse the slow, ponderous tree-creature Treebeard to action in The Lord of the Rings. The comparison, intended to capture the gap between fast-moving technology and slow-moving regulation, took on an unexpected resonance just days later.

This is a fast-moving and developing story, with Anthropic reportedly set to meet with the administration over the dispute. The coming days are likely to bring further clarity on whether access is restored and how the broader standoff is resolved.