U.S. Blocks Anthropic's Fable 5 and Mythos 5 from Foreign Users Due to Export Controls
By directive from the U.S. Department of Commerce, Anthropic has suspended access to the Fable 5 and Mythos 5 models for foreign nationals. The company disabled both for its entire customer base on the night of June 13, less than 96 hours after launch.
Four Days of Existence
Launched on June 9 with a context window of 1 million tokens, output of up to 128,000 tokens per request, and a price of $10 per million input tokens and $50 per million output tokens, the Fable 5 was the most capable model that Anthropic had made available to the public. The company described it as belonging to a new class above Opus, named Mythos, with Mythos 5 restricted to selected partners. On the night of Friday, June 13, both models were taken offline for all customers worldwide.
U.S. Secretary of Commerce Howard Lutnick sent a letter to Anthropic CEO Dario Amodei directing that the Fable 5 and Mythos 5 be subject to export controls to any location outside the U.S. and to all foreign nationals within American territory, including employees of Anthropic without U.S. citizenship. To ensure compliance before implementing any citizenship verification system, Anthropic chose to deactivate both models for its entire customer base.
The Trigger: A Mythos Jailbreak
The motivation for the order, according to government sources interviewed by Axios and confirmed to Bloomberg, was a report by an unidentified company to the Department of Commerce claiming to have circumvented the safeguards of the Mythos 5. The model was noted for its ability to identify vulnerabilities in software, raising alarms about potential offensive use in critical infrastructure.
In an official statement published on the company's website, Anthropic contested the premise of the order. According to the statement, "our understanding is that the government believes it has become aware of a bypass method for the Fable 5" and that "the level of capability shown is widely available in other models, including OpenAI's GPT-5.5." The company warned that if the same threshold is uniformly applied by the government, the outcome would be the halt of all new frontier model releases in the U.S., not just its own.
The Background with the Pentagon
The order did not emerge in isolation. In the previous months, the Department of Defense had designated Anthropic as a risk to the military supply chain after the company refused the "all lawful use" standard required for deployment in defense systems. Anthropic has maintained restrictions on mass domestic surveillance and autonomous weapons, positions that put it on a collision course with the Trump administration. The combination of the Pentagon's ban with the Department of Commerce's export control places Anthropic in an unprecedented regulatory position for an American AI company: banned by its own government and blocked from serving customers abroad.
The Impact Beyond U.S. Borders
The order affects anyone without U.S. citizenship, regardless of location. For large European consultancies, the impact is immediate and operational. Capgemini, based in Paris, and Accenture, with centers of excellence in Ireland and Germany, integrate Claude models into their software development pipelines via Claude Code. These teams, predominantly composed of European Union citizens, lost access to Fable 5 and Mythos 5 on the night of June 13.
In India, the impact is structural. TCS, Infosys, and Wipro have more than 1.2 million IT professionals, almost all Indian citizens. The three companies use Anthropic's API in code automation pipelines and have supply contracts based on Claude models. A technical migration to OpenAI’s GPT-5.5 or Google’s Gemini 3.5 Pro may be feasible in weeks, but contracts with clients specifying Claude models create an operational liability that is more difficult to absorb in the short term.
For companies outside the U.S., the export control materializes a risk that governance experts have been calling a reliance on a single model in an asset that can be regulated as strategic infrastructure. The rule currently covers Fable 5 and Mythos 5. The regulatory logic need not stop there.
The Precedent
This marks the first documented instance of a national government applying export controls to an AI model specifically based on its offensive capabilities in cybersecurity. GPU controls regulate hardware. This order regulates the intelligence embedded in the model, creating a legal framework that regulators in the European Union and Japan will closely observe, especially as the European AI Office prepares its framework for evaluating general-use models, with a start date expected in August 2026.
Anthropic had not announced whether it plans to contest the order legally by the time this article was published.