Anthropic Halts Fable 5 and Mythos 5 After US Export Control Directive

Anthropic shut down access to its Fable 5 and Mythos 5 models over the weekend after the US government issued an export control directive citing national security authorities. The order, which applies to any foreign national inside or outside the United States, forced Anthropic to disable the models for all customers to ensure compliance.

The directive was driven in part by fears that Mythos had been accessed by a group linked to China, according to a report from Semafor cited by The Verge. If the Chinese government had access to either model, it would present a serious national security concern.

Disruption to workflows and business continuity

Users who had built workflows around Fable 5 or Mythos 5 found those workflows stopped abruptly. One developer reported that a workflow making solid progress with Fable 5 was halted, and switching to Opus 4.7 or 4.8 resulted in much lower accuracy. The incident highlights a practical risk for businesses that depend on a single frontier model for critical tasks.

• Anthropic said the order was based on a report that it believes was the basis for a universal jailbreak, a capability also offered by other models.
• The company argued that if the same logic were applied to all frontier models, it could halt new deployments of frontier models entirely.
• Canadian Prime Minister Mark Carney compared the shutdown to the 2008 financial crisis, warning of systemic vulnerability from depending on a small number of powerful AI models.
• In India, Anthropic's second-largest market, the shutdown strengthened arguments for sovereign AI infrastructure, as reported by The Next Web.

Implications for AI infrastructure and sovereignty

The shutdown is a wake-up call for the importance of local AI serving, especially for critical workflows. If multiple models like Fable 5, Mythos, GPT-5.5, or their successors are all taken offline simultaneously, businesses with no fallback could face significant operational risk. The incident also bolsters the case for countries like India to invest in their own AI models and infrastructure, reducing dependence on US-based providers.

What comes next is uncertain. Anthropic has not indicated when or if the models will be restored. The broader industry may need to reconsider how it builds and deploys frontier models, with an eye toward resilience and export control compliance. For now, the message is clear: relying on a single AI model for critical workflows carries real, and potentially sudden, risk.