Trump administration lifts export controls on Anthropic's Fable 5 and Mythos 5

The Trump administration has quietly restored Anthropic's most powerful AI models after three weeks off the market.

Commerce Department lifts the export controls

The US Department of Commerce told Anthropic on Tuesday 30 June 2026 that its Bureau of Industry and Security had withdrawn the export controls on Claude Fable 5 and Claude Mythos 5, ending a standoff that had taken the company's two newest models off the market for roughly three weeks.verifiedVerified Sourced from Anthropic says Trump administration lifted restrictions on some of its most powerful Claude AI models. Commerce Secretary Howard Lutnick wrote that 'over the past two weeks, we have worked closely with Anthropic to analyze and approve Fable 5 to ensure alignment across the US Government and strengthen America's leadership in AI'.verifiedVerified Sourced from Anthropic says Trump administration lifted restrictions on some of its most powerful Claude AI models.

Anthropic said on X that Fable 5 would return globally on Claude.ai and Claude Code from Wednesday 1 July, capped at 50 per cent of existing weekly usage limits until 7 July, and would be re-enabled on Amazon Web Services, Google Cloud and Microsoft Foundry as soon as possible.verifiedVerified Sourced from White House Lifts Restrictions On Anthropic's Mythos And Fable AI Models. Mythos 5, the more capable variant, has been restored only to a set of US organisations approved by the government to defend critical infrastructure .

How the ban started

Anthropic launched Fable 5 and Mythos 5 on 9 June 2026, describing Fable as a 'Mythos-class model made safe for general use' with new classifiers that block cybersecurity, biology, chemistry and model-distillation requests. Three days later, on the evening of Friday 12 June, Lutnick sent CEO Dario Amodei a directive requiring the company to suspend access for all foreign nationals worldwide, including foreign-national employees inside Anthropic itself. Anthropic chose to disable both models rather than attempt to enforce the restriction on its user base.

White House AI adviser David Sacks said on X the trigger was a jailbreak technique that Amazon chief executive Andy Jassy had flagged to the administration, and that 'the Admin asked Dario to fix the jailbreak or de-deploy the model. Dario refused'. Anthropic countered that the technique was 'a narrow one that would unlock Mythos's cybersecurity capabilities in only one specific instance' and said the capability was already available in other public models. Semafor reported that officials were also concerned a China-linked group may have accessed Mythos.

What Anthropic agreed to

In exchange for the export licence being lifted, Anthropic agreed to pre-release future frontier models to federal reviewers for testing, work with the government on shared jailbreak-severity scoring and vulnerability-response protocols, and help develop an industry-wide safety framework with Amazon, Google and Microsoft. Anthropic also said it had implemented a new safeguard that blocks the specific jailbreak the government objected to in about 99 per cent of tested cases.

The company said in a statement that its 'hope is that this collaboration, along with our proposed consensus industry framework, will serve as the basis for systematic rules for the whole industry', and thanked users on X for their patience 'and to everyone who worked with us on redeploying the models' .

Cybersecurity researchers pushed back

The White House's decision to use export controls to force a change on a US commercial AI product drew rare public criticism from cybersecurity researchers. Katie Moussouris told CyberScoop the disputed capability was 'not a guardrail bypass. It is the most valuable thing an AI model can do for defensive security', and called the controls 'heavy handed' and 'misguided'. An open letter signed by dozens of researchers argued Fable 5's guardrails were 'notoriously oversensitive' compared with rival models and that the flagged behaviour was not uniquely dangerous.

Democratic Senator Mark Warner of Virginia said the administration's approach lacked a 'transparent, risk-based process with clear rules and consistent standards' and questioned whether the decision reflected 'objective national security concerns or something else'.

The Australia angle

Anthropic opened a Sydney office in April 2026, the company's fourth in the Asia-Pacific after Tokyo, Bengaluru and Seoul, with Theo Hourmouzis appointed general manager for Australia and New Zealand . Australian customers named by Anthropic include Commonwealth Bank, Canva, Quantium, Xero and YMCA South Australia, along with research partners Australian National University, Murdoch Children's Research Institute, the Garvan Institute of Medical Research and Curtin University . Anthropic says Australia and New Zealand rank fourth and eighth globally for Claude.ai usage relative to population .

Francesco Bailo, deputy director of the AI, Trust and Governance Centre at the University of Sydney, told Al Jazeera the US government had likely 'overreacted' and that blocking the models could set problematic regulatory precedents affecting other AI developers .