Anthropic cuts off Fable 5 and Mythos 5 after US export directive
Anthropic said a Commerce Department order forced it to disable two new AI models for all customers while it works through compliance concerns.
By Hana Yoshida · Markets Reporter
3 min read
Anthropic shut down customer access to its newly released Fable 5 and Mythos 5 AI models on Friday night after receiving a directive from the US Commerce Department. The company said the order put the models under export controls that restrict their use outside the United States, leaving it unable to keep them available while ensuring immediate compliance.
In a Friday night notice, Anthropic said it had disabled both models for all customers because it could not quickly limit access in a way that satisfied the government order. The company said access to its other models was unchanged.
The shutdown came only days after Fable 5 and Mythos 5 launched. Anthropic apologized to customers for the disruption and said it planned to release more information within 24 hours.
Axios reported that a Trump administration official said the government was concerned about reports of a jailbreak affecting Fable 5. According to Axios, the alleged workaround could bypass broad classifier-based safeguards intended to block prompts involving cybersecurity, chemistry and biology.
Axios reported that the administration asked for a pause in the models’ release to give national security agencies time to strengthen protections against the type of threat officials believed the models could pose. The official cited by Axios said that work could be finished in the next few weeks.
Anthropic disputed the scale of the concern in its own statement. The company said the government had supplied only verbal evidence of what it described as a narrow and non-universal jailbreak involving a request for Fable 5 to examine a specific codebase for software flaws.
Anthropic said the evidence it had seen showed the technique being used to identify minor and relatively simple software vulnerabilities. The company also said other public models, including GPT-5.5, have comparable capabilities in that area.
While Anthropic said it would comply with the legal directive, it argued that the reported issue did not justify recalling a commercial model already available to a large user base. The company said applying the same standard across AI developers would, in its view, effectively stop new releases by frontier model providers.
The dispute follows President Donald Trump’s executive order earlier this month urging AI model makers to submit voluntarily to government security testing. The signing followed an earlier postponed ceremony, after reported internal administration disagreements over the order.
For Anthropic customers, the practical result is immediate: Fable 5 and Mythos 5 are offline, with no public timetable for restoration. Anthropic framed the move as a temporary compliance step tied to a government directive, while the administration concern reported by Axios centers on whether safeguards around advanced AI models are strong enough before broad release.
This story draws on original reporting from Ars Technica.