Anthropic's Mythos 5 gets limited carveout from US restrictions

June 28, 2026 - 00:46

Anthropic has announced it will bring back access to its Mythos 5 model for a specific group of US entities. The company stated it is restoring the model for organizations that "operate and defend critical infrastructure." This move comes as a limited carveout from broader US restrictions that had previously limited the model's availability.

The decision targets sectors like energy, water, and transportation, where the company sees a need for advanced AI tools to protect essential systems. Anthropic emphasized that the exception is tightly controlled and does not apply to general commercial or public use. The company also noted that it will monitor how these organizations use Mythos 5 to ensure compliance with safety guidelines.

Critics have questioned whether this selective access could create security gaps, but Anthropic argued that the benefits of using the model for threat detection and response outweigh the risks. The company is working with federal agencies to define the boundaries of this exception. For now, only verified infrastructure operators will be able to use the model, with strict oversight on data handling and deployment. This marks a shift in how AI companies navigate government restrictions while still supporting national security priorities.