Anthropic Suspends Fable 5 and Mythos 5 Following Government Export Directive

The U.S. government has ordered Anthropic to disable its Fable 5 and Mythos 5 models, citing national security concerns following cybersecurity research shared by Amazon.

AI & Culture Desk
Updated Nov 8, 2026 3 min read
Anthropic Suspends Fable 5 and Mythos 5 Following Government Export Directive

On Friday evening, Anthropic was ordered by the U.S. government to immediately suspend access to its Fable 5 and Mythos 5 models, according to reports from The Verge and TechCrunch. The directive, issued at 5:21 p.m. ET, was framed as an export control action intended to restrict foreign nationals from accessing the systems. However, the company opted to take the models offline globally for all users, including its own employees, to ensure compliance with the broad federal mandate. This marks a significant escalation in the regulatory environment for frontier AI developers, effectively grounding the company's most capable technology.

The intervention appears to be the result of cybersecurity research conducted by Amazon, a major financial backer and cloud provider for Anthropic. According to The Wall Street Journal, Amazon CEO Andy Jassy engaged in discussions with the White House regarding the findings shortly before the directive was issued. The research reportedly demonstrated that Fable 5 could be prompted to provide information useful for cyberattacks. Amazon has not publicly commented on these reports. This development highlights the complex, often adversarial nature of the relationships between hyperscalers and the AI startups they fund, where cloud giants may prioritize government relations over the continued operation of their partners' products.

Anthropic has publicly disputed the government's characterization of the situation. In a statement, the company noted that it had not received specific details regarding the national security concerns, claiming that evidence of the alleged jailbreak was provided only verbally. Anthropic argued that the vulnerabilities identified are minor, non-universal, and consistent with capabilities already present in other publicly available models, such as OpenAI’s GPT-5.5. Furthermore, the company asserted that its independent classifier systems remain effective at preventing harmful outputs, even if a user attempts to bypass standard guardrails.

The tension between Anthropic and the current administration is not new. According to The Verge, the two parties have clashed previously over the company’s refusal to facilitate mass surveillance or support lethal autonomous weapons. In February, the administration instructed federal agencies to cease using Anthropic’s models, and the Department of Defense briefly labeled the company a supply chain risk. While the two sides had seemingly reconciled to expand access to Mythos, this latest order suggests that the relationship remains volatile. Observers note that Anthropic’s own marketing—which emphasized the extreme power and potential danger of the Mythos model—may have inadvertently invited this level of scrutiny.

The shutdown specifically impacts the Fable 5 model, which was released just three days prior to the order, and the Mythos 5 model. Mythos had been kept under strict access controls via Project Glasswing, a program that shared the model with roughly 50 vetted organizations, including Google, Microsoft, and CrowdStrike, for defensive cybersecurity research. Fable 5 was intended to be the commercially available version, equipped with guardrails designed to block high-risk queries. Anthropic warned that applying such a stringent standard for model deployment would likely halt all new releases across the entire frontier AI industry.

The incident underscores the regulatory precarity of relying on centralized AI APIs. As Anthropic prepares for a potential IPO this year, the sudden blackout serves as a stark reminder of the sovereign risks inherent in the current AI ecosystem. While some security researchers, such as Luta Security CEO Katie Moussouris, have questioned the severity of the alleged jailbreaks, the government’s action has nonetheless succeeded in forcing a total service suspension. For developers and enterprise customers, the event is likely to accelerate interest in multi-model redundancy and open-source alternatives to mitigate the risk of sudden regulatory deplatforming.

Sources

  1. 01 Anthropic cuts off Fable 5 and Mythos 5 access following government order — The Verge
  2. 02 Amazon security research reportedly led to the White House’s Anthropic Fable ban — The Verge
  3. 03 Anthropic’s safety warnings may have just backfired — the government has pulled the plug on its most powerful AI — TechCrunch