Anthropic Disables Fable 5 And Mythos 5 After U.S. Export Directive

Anthropic has disabled Fable 5 and Mythos 5 for all customers after a U.S. government export control directive restricted access by foreign nationals, including foreign national employees working inside the company.

The directive requires Anthropic to suspend access to the two models for any foreign national, whether inside or outside the United States. The company received the order on Friday at 5:21 p.m. ET and removed the models globally to comply. Access to other Anthropic models remains unaffected.

Commerce Secretary Howard Lutnick sent the order to Anthropic CEO Dario Amodei, placing Fable 5 and Mythos 5 under a licensing regime for export, re-export or domestic transfer involving foreign persons. That structure makes access to the models a national security and licensing issue rather than a normal commercial product rollout.

The shutdown arrived only days after Anthropic launched Claude Fable 5 and Claude Mythos 5. Fable 5 was introduced as a Mythos-class model configured for general use, while Mythos 5 was positioned as a restricted version for vetted cyberdefenders and infrastructure partners.

Jailbreak Concern Drives The Dispute

The technical dispute centers on a reported method for bypassing Fable 5’s safeguards. Anthropic said the demonstration it reviewed involved a narrow technique used to identify a small number of previously known, minor vulnerabilities. The company also said other publicly available models can find the same types of issues without the alleged bypass.

Anthropic’s launch structure had already separated the two models by risk controls. Fable 5 was designed with safeguards that route sensitive cybersecurity, biology and other high-risk requests away from the Mythos-class model. Mythos 5 used the same underlying model but with some safeguards lifted for approved defensive users.

That distinction was central to the public Mythos-class release. Fable 5 gave ordinary users access to a high-capability model with guardrails, while Mythos 5 remained tied to Project Glasswing and trusted-access use cases.

The new directive overrides that commercial split. Anthropic is complying with the order while disputing the basis for the recall, arguing that a narrow potential jailbreak should not force the removal of a model already deployed at scale.

AI Security Debate Hits Developers And Crypto Teams

The shutdown immediately affects developers, enterprise users and teams that had started testing Fable 5 for coding, research, analysis, long-context work and agent-style workflows. Foreign employees and foreign customers are included in the restriction, which pushed Anthropic toward a full customer cutoff instead of a country-by-country access block.

The decision also revives the security debate around frontier AI models and vulnerability discovery. Earlier concerns around unauthorized Mythos Preview access had already put AI-assisted exploit discovery, model containment and defensive access programs under pressure.

For crypto teams, the issue sits close to wallet security, smart-contract auditing, infrastructure hardening and AI-assisted code review. Stronger models can help defenders find flaws faster, but they also raise the risk that weak approvals, old dependencies, exposed keys, insecure front ends and neglected contracts become easier to probe.

The order adds a regulatory layer to Anthropic’s broader growth story. The company has been moving toward public-market scrutiny after its confidential IPO filing, while Claude’s most advanced models are now caught between enterprise demand, national security review and developer access.