U.S. Lifts Export Controls on Anthropic's Most Advanced AI Models After 18-Day Standoff

Lynn Matthews | WECU Media

WASHINGTON — The U.S. Commerce Department has withdrawn export controls on Anthropic's two most advanced artificial intelligence models, Claude Fable 5 and Claude Mythos 5, ending an 18-day standoff that had knocked the company's flagship systems offline worldwide.

Commerce Secretary Howard Lutnick announced the decision Tuesday, saying in a letter to Anthropic that a license is “no longer required for the export, reexport, or in-country transfer, including deemed export or deemed reexport, of the Mythos or Fable models.”

Anthropic confirmed the reversal within hours in a post on X: “We've received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5. We'll begin restoring access tomorrow, and will share an update soon.”

Access to Fable 5 began returning globally Wednesday across the Claude platform, Claude.ai, Claude Code and Claude Cowork, though usage will be capped at up to 50% of normal weekly limits through July 7 before full availability resumes, the company said in a newsroom post titled “Redeploying Claude Fable 5.”

How the Anthropic Freeze Started

The Commerce Department's Bureau of Industry and Security ordered Anthropic on June 12 to cut off access to both models for any foreign national — including the company's own noncitizen employees — whether inside or outside the United States.

Anthropic said it was given roughly 90 minutes to comply and, lacking a reliable way to verify user nationality in real time, disabled both models for all users worldwide.

The trigger, Anthropic said in its newsroom post, was a report that researchers at Amazon had found a way to bypass Fable 5's safeguards — a “jailbreak” — that led the model to identify software vulnerabilities and, in one case, produce code demonstrating how one of those vulnerabilities could be exploited. The Wall Street Journal reported, per 9to5Mac, that conversations between Amazon Chief Executive Andy Jassy and White House officials helped prompt the export-control directive.

Anthropic disputed that Fable 5 posed a unique risk. The company said its testing found that less capable models — including its own Claude Opus 4.8, along with OpenAI's GPT-5.5 and Moonshot AI's Kimi K2.7 — could identify the same vulnerabilities Fable 5 had flagged. Every model Anthropic tested, including older versions such as Claude Haiku 4.5 and Sonnet 4.6, was able to produce the same exploit demonstration, the company said in its newsroom post.

A Staged Rollback

The government's position eased in stages. On June 26, Lutnick authorized a limited release of Mythos 5 — the more powerful, less-restricted model underlying the public-facing Fable 5 — to what Anthropic described as “a small group of cyber defenders and infrastructure providers.” Fable 5 remained blocked until Tuesday's full withdrawal of the controls.

In his letter, Lutnick said Anthropic had agreed to “proactively detect and address security risks associated with the models,” to work with the government on release protocols and standards for the Fable and Mythos models “and future models,” and to inform the government of any malicious activity involving them. Commerce reserved the right to reimpose licensing requirements if Anthropic fails to meet those commitments or if circumstances change.

Broader Tensions

The export-control dispute followed months of friction between Anthropic and the Trump administration. In February, Trump ordered federal agencies to stop using Anthropic's technology amid a dispute with the Pentagon over safeguards for military applications; the Defense Department had designated Anthropic a supply-chain risk. Anthropic sued in March after declining to change its usage policies to grant the government the access it sought, citing concerns the technology could be used for mass surveillance or fully autonomous weapons.

Reaction

The blanket shutdown drew criticism from researchers and industry figures who argued the restrictions overstated the risk and cost the U.S. competitive ground. Francesco Bailo, deputy director of the AI, Trust and Governance Centre at the University of Sydney, told Al Jazeera the restoration had been expected in tech circles, since reports of the Fable 5 jailbreak had been “widely inflated beyond their actual significance.”

Tanishq Abraham, a former research director at Stability AI who now leads the medical AI company Sophont, called the reversal “a big deal” that raises questions about how the industry will be regulated going forward. “The biggest question now is: What precedent does this set for the industry? Does the U.S. government need to approve every frontier model release?” Abraham told Al Jazeera.

That question isn't hypothetical. OpenAI said last week it would roll out its newest model series, GPT-5.6, to “a small group of trusted partners” first, following pressure from the U.S. government to stagger the release — a sign the Fable 5 episode may be shaping how the broader industry handles future launches.

What's Next

Commerce's letter, dated June 30, was addressed to Anthropic's chief compute officer, Tom Brown, and cited the agency's updated evaluation of “diversion risk” for both models as the basis for withdrawing the restrictions entirely — a detail reported by Cyber Security News but not yet independently confirmed elsewhere.

Anthropic said in a follow-up post that it is redeploying Fable 5 “with a new set of classifiers to target and block more cybersecurity tasks” following what it called “productive” conversations with the government.

Sources: Anthropic newsroom; Anthropic on X; Fox Business; 9to5Mac; Al Jazeera; Cyber Security News; The Epoch Times; CNN Business; Forbes.