SAN FRANCISCO — Marking an unprecedented escalation in federal artificial intelligence regulation, Anthropic announced the immediate worldwide suspension of its most advanced AI models, Claude Fable 5 and Mythos 5.
The sudden move comes after the US Department of Commerce issued a sweeping export control directive on June 12, 2026, ordering the company to block all foreign nationals from accessing the systems due to immediate national security concerns.
The Anatomy of a Global Shutdown
The directive targeted access by any foreign national, whether located inside or outside the United States. Crucially, the mandate extended to Anthropic’s own foreign-national employees, forcing the firm to take a drastic operational step:
-
Total Recalibration: Because isolating users based purely on nationality across large-scale commercial cloud infrastructures is functionally impossible without risking non-compliance, Anthropic chose to disable both models for all customers globally.
-
Isolated Disruption: The company verified that access to its other mainstream Claude models (such as Opus, Sonnet, and Haiku variants) remains completely unaffected and operational.
-
The User View: Subscribers attempting to leverage the “Mythos-class” system are currently greeted with a “Currently Unavailable” notice on the Claude interface, while Amazon Web Services (AWS) confirmed it revoked access across all hosting regions at Anthropic’s request.
Security Clashes: Cybersecurity Capabilities vs. “Jailbreak” Risks
The regulatory intervention took place just three days after Anthropic publicly launched Fable 5, which was specifically engineered for advanced mathematics, visual reasoning, and highly optimized software coding.
| Model Variable | System Architecture & Launch Footprint |
| The Underlying Tech | Both systems run on Anthropic’s “Mythos-class” architecture, a high-capability tier sitting above older Opus models, designed to flag and fix complex software vulnerabilities. |
| The Guardrails | Fable 5 launched publicly with filters blocking highly sensitive cyber queries. Mythos 5 was restricted to vetted cyber experts under “Project Glasswing.” |
| The US Gov Alarm | Regulators were allegedly spooked by an Amazon research report detailing a narrow, non-universal jailbreak where prompts successfully bypassed filters to exploit small software flaws. |
Anthropic pushed back sharply against the Commerce Department’s intervention, labeling the shutdown a “misunderstanding.” In an official statement, the AI firm argued:
“We reviewed a report that we believe is the basis of the government’s directive and validated that the level of capability displayed there is widely available from other models (including OpenAI’s GPT-5.5), and is used every day by the defenders who keep systems safe… We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people.”
Geopolitical Friction and Pre-IPO Tensions
The export control order represents a fundamental shift in AI regulation. Previously, Washington restricted the physical hardware and semiconductor chips (like Nvidia GPUs) used to train models; this order marks the first time the US government has restricted the software deployment of a finished model as a national security asset.
The enforcement lands amidst an ongoing legal dispute between Anthropic and the administration. Earlier in 2026, the Pentagon designated Anthropic a “supply chain risk” after the company refused to let its models be integrated into fully autonomous weapons systems or domestic surveillance operations.
The abrupt technical recall also creates a highly turbulent financial environment for Anthropic, arriving just as the company prepares for an initial public offering (IPO) amid soaring market valuations for its core aerospace and AI competitors.
