Anthropic has disabled access to its two most advanced AI models globally after the United States government issued an export control directive requiring it to suspend access for all foreign nationals just three days after it became publicly available.
The suspension of Claude Fable 5 and Claude Mythos 5, in force since 12 June 2026, marks the first time the US government has applied export controls directly to an AI software model rather than to the chips or hardware used to run it.
Anthropic launched Claude Fable 5 on 9 June 2026, describing it as the most capable AI model it had ever made generally available. Fable 5 is a consumer and enterprise release of what Anthropic calls a "Mythos-class" model, its highest internal capability tier. Claude Mythos 5 itself had been separately available only to a small number of vetted partners through Anthropic's restricted Project Glasswing programme, due to concerns about its capabilities in areas including cybersecurity, biology, and chemistry.
Anthropic said at launch that releasing a model of Fable 5's capability "comes with risks", and said it had introduced safeguards routing certain sensitive queries, particularly those involving offensive cybersecurity, advanced biology, and chemistry, to Claude Opus 4.8, its next most capable publicly available model, rather than processing them through Fable 5 directly.
On 12 June, the US Commerce Department sent a letter to Anthropic's CEO Dario Amodei ordering the company to immediately suspend access to both Fable 5 and Mythos 5 for any foreign national, whether located inside or outside the United States. The directive explicitly covered Anthropic's own foreign national employees.
Anthropic said the letter did not provide specific details of the government's national security concern, but that its understanding was that the government believed it had identified a method of bypassing ("jailbreaking") Fable 5's safeguards. The company said it reviewed a demonstration of the technique and concluded it amounted to asking the model to read a specific codebase and identify software flaws, producing results that were "previously known, minor vulnerabilities" also discoverable using other publicly available models including OpenAI's GPT-5.5.
Unable to verify user nationality at the API level in real time, Anthropic disabled both models for all customers globally rather than attempt selective filtering. AWS, Google Cloud, Microsoft Foundry, and other platforms routing traffic through Anthropic's API executed simultaneous takedowns.
Anthropic complied with the directive but publicly disputed its basis. In its official statement the company said it disagreed that a narrow, non-universal jailbreak should be grounds for recalling a model deployed to hundreds of millions of people, and argued that applying such a standard across the industry "would essentially halt all new model deployments for all frontier model providers." It added that it had not received a disclosure of any jailbreak that had produced a harmful result, and that the government had provided only verbal evidence of the technique.
The company said it believed the action was "a misunderstanding" and that it was working to restore access "as soon as possible." It did not provide a timeline.
The practical effect of the directive has been felt immediately and broadly outside the United States. UK organisations using Fable 5 or Mythos 5 through the Claude API, AWS Bedrock, or other cloud platforms lost access without warning. Project Glasswing partners, organisations that had been vetted for access to Mythos 5 specifically for defensive security research, in nations including the UK and South Korea, were also cut off.
The UK's Minister for AI and Online Safety, Kanishka Narayan, responded by framing the incident as evidence of the need for technological sovereignty, pointing to the government's £1.1 billion investment in domestic AI chip infrastructure. The episode has renewed debate about the extent to which public sector organisations and critical infrastructure providers in allied nations can rely on US-hosted frontier AI services without risk of abrupt access withdrawal.
