Statement on the US government directive to suspend access to Fable 5 and Mythos 5
Anthropic has suspended access to its Fable 5 and Mythos 5 models globally, following a US government directive citing national security concerns over a potential "jailbreak" method.
Intelligence analysis by Gemini 2.5 Flash
Anthropic was ordered by the US government to immediately suspend access to its Fable 5 and Mythos 5 AI models for all foreign nationals due to national security concerns. The company complied, resulting in a global shutdown of these models, but expressed disagreement with the decision, stating the identified vulnerability is minor and common across the industry.
Imagine a super-smart computer program that helps people, but it has some secret rules to keep it from doing anything bad. The government found a tiny trick, like a secret handshake, that let someone get around one of these rules, even though it only worked for small things and other programs could do the same trick. Because of this, the government made the company turn off the program for everyone, which the company thinks is a bit unfair for such a small trick.
Analysis
On June 12, 2026, Anthropic announced the immediate suspension of global access to its Fable 5 and Mythos 5 AI models. This action was a direct result of an export control directive issued by the US government, which cited national security authorities and mandated the suspension of access for all foreign nationals, including Anthropic's own foreign national employees.
The government's directive, received at 5:21 PM ET on the day of the announcement, did not provide specific details regarding the national security concern. However, Anthropic understands that the government believes a method exists to bypass, or "jailbreak," Fable 5. Anthropic reviewed a demonstration of this technique, noting it exploited a small number of previously known, minor vulnerabilities. The company asserts that these vulnerabilities are relatively simple and can also be discovered using other publicly available models.
Anthropic emphasized its "defense in depth" strategy for Fable 5, which includes strong safeguards against misuse in areas like cybersecurity. They highlighted thousands of hours of red-teaming with government agencies and third-party organizations prior to launch, which showed Fable's safeguards to be more effective than previous models. The company stated that no testers had found a "universal jailbreak"—a method that could broadly bypass safeguards and unblock a wide range of cyber capabilities. They concede that perfect jailbreak resistance is likely impossible for any model today and that non-universal jailbreaks (which elicit specific cyber information) are a known industry vulnerability.
To combat these risks, Anthropic implemented a policy of 30-day customer data retention for Fable, enabling them to research and mitigate jailbreaks. They believe this strategy makes Fable's risks comparable to other deployed industry models. Anthropic claims they have not received disclosures of concerning non-universal jailbreaks leading to harmful results; disclosed issues were either benign or minor and offered no Mythos-specific uplift.
Despite complying with the legal directive, Anthropic disagrees with the government's decision to recall a commercial model deployed to hundreds of millions over a narrow, non-universal jailbreak. They argue that applying such a standard across the industry would effectively halt all new frontier model deployments. Anthropic advocates for a transparent, fair, clear, and technically grounded statutory process for blocking unsafe deployments, which they believe this action did not adhere to. The company apologized to customers for the disruption and is working to restore access.
Key points
- Anthropic suspended access to Fable 5 and Mythos 5 globally due to a US government directive.
- The directive cited national security concerns over a "jailbreak" technique for Fable 5.
- Anthropic claims the identified vulnerabilities are minor and accessible via other public models.
- The company disagrees with the recall, arguing it sets an unsustainable precedent for AI model deployment.
- Anthropic is working to restore access and advocates for a transparent and technically grounded regulatory process.
Anthropic is actively working to restore access to Fable 5 and Mythos 5, indicating a potential resolution and reinstatement of these models. The company's commitment to a 'defense in depth' strategy suggests continued efforts to enhance AI safety while also pushing for transparent regulatory processes, which could lead to more robust and clear guidelines for future AI deployments.
The US government's directive, if applied broadly, could significantly hinder the deployment of new frontier AI models across the industry, leading to a chilling effect on innovation. The lack of specific details regarding national security concerns and a transparent regulatory process creates uncertainty and could lead to ongoing conflicts between AI developers and government oversight.
