Anthropic says it has disabled access to its Fable 5 and Mythos 5 AI models after receiving a US government export-control directive tied to national security concerns.
The company said the directive required it to block access by foreign nationals, including people outside the United States and foreign national employees working inside the company. Anthropic said it responded by disabling the models for customers more broadly while it works through compliance.
Access to earlier Anthropic models, including Opus and Sonnet, was described as unaffected by the order.
What Anthropic says happened
Anthropic said it received a government letter directing the company to restrict access to Fable 5 and Mythos 5. The company said the notice cited national security authorities but did not give detailed technical reasoning for the decision.
Because the restriction covered foreign nationals regardless of location, Anthropic said it could not simply apply the order by geography. The company presented the wider shutdown as the practical result of complying with the directive while avoiding accidental access by restricted users.
The source material describes Fable 5 as the broader-use model and Mythos 5 as a more tightly controlled model aimed at specialized work, but those product descriptions should be treated cautiously unless confirmed directly by Anthropic. What is clear from Anthropic’s statement is that both models were subject to the same access action.
The move puts AI export controls back in the center of the policy debate around frontier models, especially when a system’s cyber or scientific capabilities become part of a national security assessment.
The suspected concern: model jailbreaks
Anthropic said it believes the government’s concern may relate to a possible way to bypass, or jailbreak, safeguards in Fable 5. The company did not describe a confirmed universal jailbreak, and the available account does not support treating the suspected vulnerability as proven public fact.
Instead, Anthropic argued that any potential issue appeared limited and not unique enough to justify withdrawing a commercial model from broad use. The company also said it had spent substantial time on safety testing and red-team work with government, outside, and internal reviewers before release.
Those safety claims remain Anthropic’s position. They should not be read as independent proof that the models have no exploitable weaknesses. The more careful reading is that Anthropic disputes the government’s apparent risk assessment and says it has not been shown evidence of a broad, harmful bypass that would warrant such a sweeping restriction.
Why the order matters beyond Anthropic
The immediate effect is straightforward: customers who expected to use Fable 5 or Mythos 5 lost access while Anthropic works through the directive. For companies testing the models in software development, security, or research workflows, that creates a reliability problem separate from the models’ technical performance.
The broader issue is precedent. Anthropic warned that if a narrow or unresolved jailbreak concern is enough to trigger a commercial model recall, other frontier AI launches could face similar disruption. That concern is especially relevant for enterprise buyers, who may need to assess not only model quality and pricing but also regulatory exposure.
For AI vendors, the case shows how safety testing, export controls, employee access rules, and customer availability can collide quickly. A model can be technically available one day and commercially unavailable the next if regulators decide its capabilities create unacceptable risk.
What remains unclear
Several important details were not fully established in the source material. The government has not publicly laid out the specific national security concern behind the directive. The exact nature of any alleged jailbreak has not been independently verified. It is also unclear how quickly Anthropic can restore access, or whether restored access would come with new eligibility checks, geography rules, or customer restrictions.
Anthropic said it is working to restore access and plans to provide more information. Until more details are available, the safest conclusion is limited: the company says it disabled Fable 5 and Mythos 5 because of a US export-control directive, it disagrees with the basis for that action, and it is trying to resolve the restriction without affecting its other models.