National Security Directive Forces Anthropic to Suspend Access to Claude Fable 5 and Mythos 5 Following “Jailbreak” Concerns

Anthropic has abruptly disabled access to its high-performance Mythos-class AI models, Fable 5 and Mythos 5, following a US government export control directive citing national security risks. While the government alleges a potential “jailbreak” vulnerability, Anthropic maintains the models’ safeguards are industry-leading and is working to restore service for its global customer base.

RMN News Technology Desk
New Delhi | June 13, 2026

Anthropic issued an urgent statement on June 12, 2026, announcing the immediate suspension of its most advanced AI models, Claude Fable 5 and Claude Mythos 5. The move comes in response to a US government export control directive received at 5:21 p.m. ET, which prohibits access to these models by any foreign national, including Anthropic’s own international employees. To ensure full compliance with the directive, Anthropic was forced to disable access for all customers globally.

The government’s intervention stems from concerns regarding a potential method for “jailbreaking” Fable 5, which could reportedly allow users to bypass the model’s safety protocols. However, Anthropic has publicly disagreed with the severity of the findings, characterizing the alleged vulnerability as a “narrow, non-universal jailbreak” that merely asks the model to fix flaws in a specific codebase. Anthropic’s internal review suggests the capabilities displayed in the government’s report are already widely available in other frontier models, including OpenAI’s GPT-5.5, and are commonly used by cybersecurity defenders.

  A Contentious Recall: Anthropic maintains that the alleged “jailbreak” used to justify the recall is a capability already prevalent in other public models like GPT-5.5.

This sudden recall follows the highly successful launch of the “Mythos-class” tier just days earlier, on June 9. The models were positioned as a new industry benchmark, sitting above the previous Opus-class in both reasoning and autonomous workflow capabilities. Fable 5, designed for general use, recently demonstrated its power by completing a codebase-wide migration of 50 million lines of Ruby code for Stripe in a single day—a feat that typically requires a human team more than two months to achieve.

The suspension also affects Mythos 5, a specialized model deployed through “Project Glasswing” for critical cybersecurity and biomedical infrastructure. Mythos 5 had already begun transforming life sciences, accelerating drug design by tenfold and becoming the first model to consistently generate novel scientific hypotheses for experimental evaluation.

Anthropic remains firm in its “defense in depth” safety strategy, noting that Fable 5 was red-teamed for thousands of hours by organizations including the UK AI Safety Institute (AISI) prior to release. The company argues that if the government’s current standard for recall were applied industry-wide, it would essentially halt the deployment of all new frontier AI models. For now, access to all other Anthropic models remains unaffected while the company seeks a transparent resolution with federal authorities.