IA · 14 June 2026 · 3 min read

US Government Shuts Down Claude Fable 5 and Mythos 5 Over National Security Concerns

In brief: Anthropic has suspended global access to its frontier models, Fable 5 and Mythos 5, following a US federal export control directive. Triggered in part by an Amazon security report pointing to cyberattack risks, the shutdown exposes growing tensions between the US government and leading AI developers over regulatory authority and national security.

by Team Mocchi's

The Federal Directive and Immediate Shutdown

The global artificial intelligence sector was rocked by an abrupt regulatory action when the United States government issued an export control directive against Anthropic. As reported by Wired, the federal administration cited urgent national security concerns, requiring the San Francisco-based firm to block all foreign nationals, both inside and outside the United States, from accessing its most advanced models, Fable 5 and Mythos 5.

The scope of the order was exceptionally broad, barring even Anthropic's own foreign-born employees from using or working on the systems they helped create. Faced with the technical and legal challenges of enforcing such a precise filter across its infrastructure, Anthropic chose to take the models completely offline for all customers worldwide, shutting down the dedicated cloud servers for these two specific model families.

Amazon's Role and the Roots of the Crackdown

According to an investigation by the Wall Street Journal, detailed by The Verge, the White House's sudden intervention was partly triggered by cybersecurity research conducted by Amazon, which is ironically one of Anthropic's primary financial backers and cloud partners. Amazon CEO Andy Jassy reportedly shared these findings directly with senior administration officials shortly before the ban was finalized.

Amazon's research claimed that, through specialized and highly complex prompting techniques, Fable 5 could be coerced into bypassing its core safety filters. Once breached, the model allegedly outputted actionable instructions, system schemas, and code that could assist in executing cyberattacks on critical infrastructure. Following Jassy’s briefing, government agencies moved swiftly to enforce strict export restrictions, fearing these capabilities could be leveraged by hostile foreign states.

Anthropic's Pushback and Security Expert Skepticism

Anthropic issued a public statement complying with the order, though it openly disputed the government’s reasoning and the severity of Amazon's findings. The company, which was founded as a safety-focused Public Benefit Corporation by former OpenAI researchers Dario and Daniela Amodei, noted that federal officials had not provided specific technical details regarding the national security threat. Anthropic claimed that any potential jailbreaks presented to them were minor and did not offer any unique capabilities compared to existing systems.

Furthermore, Anthropic argued that the vulnerabilities identified are not unique to Mythos or Fable, pointing out that comparable behaviors can be replicated in other widely available commercial systems, including OpenAI's GPT-5.5. Independent cybersecurity experts have echoed some of Anthropic's skepticism. As reported by The Verge, Katie Moussouris, founder and CEO of LutaSecurity, stated publicly that she had reviewed the research paper and characterized the findings as minor, noting they did not represent a genuine jailbreak or pose a novel security risk.

Geopolitical Implications and Future AI Governance

This development marks a significant shift in relations between Washington and AI laboratories. The ban highlights a structural vulnerability for Silicon Valley tech hubs: because many top-tier AI researchers are foreign nationals, the federal export ban locked these engineers out of accessing the very models they helped build and refine.

The incident also highlights how geopolitical concerns and corporate competition are becoming deeply intertwined. The rapid federal response based on research from Amazon—a major industry competitor and partner—suggests that national security claims will increasingly impact commercial strategies and competitive dynamics in the AI landscape, forcing enterprises to carefully navigate federal oversight.

All articles on the Mocchi's blog