The US Commerce Department ordered Anthropic to halt all transfers of its most advanced AI models to non-US persons, citing national security risks
Anthropic has pulled the plug on global access to its most powerful AI models after the US government decided they were too dangerous to share with the world. The company suspended access to Claude Mythos 5, Mythos Preview, and the related Fable 5 variant as of June 2026, following a directive from the US Commerce Department.
The order came directly from Commerce Secretary Howard Lutnick in a letter to Anthropic CEO Dario Amodei on June 1, 2026. It mandated that any export, re-export, or domestic transfer of these models to non-US persons must receive governmental approval first. Anthropic publicly confirmed its compliance on June 13, 2026.
What the export controls actually require #
The suspension applies to foreign nationals located inside the United States as well. If you’re a non-US person sitting in a coffee shop in San Francisco, you’re also cut off.
The Commerce Department’s reasoning centers on the Mythos models’ advanced cybersecurity capabilities and the concern that these models could potentially be weaponized by adversaries if access isn’t tightly controlled.
Before the suspension, the Mythos models weren’t exactly available to everyone anyway. They were distributed through a program called Project Glasswing, which provided limited access primarily to US-based firms like Microsoft, Google, and NVIDIA. The focus was on addressing critical cybersecurity vulnerabilities, essentially letting America’s biggest tech companies use the models as digital immune systems.
The dual-use dilemma hits home #
The directive reflects concerns raised during the Trump administration about certain frontier AI models being too hazardous for broader foreign use. The Commerce Department’s move represents a proactive stance, intervening before any known breach or misuse rather than reacting after the fact.
What this means for investors #
The most immediate implication is that the addressable market for the most advanced AI models just got smaller. If only US persons and approved entities can access frontier cybersecurity AI, that eliminates a massive chunk of potential customers. International enterprises, foreign governments, and multinational corporations with non-US staff all face new friction.
For the broader AI industry, this sets a clear precedent. The US government is willing to treat frontier AI models the way it treats advanced semiconductors or military technology: as strategic assets that require export licensing. Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our