Early Users of Anthropic Mythos Still Retain Access After US Order
The continued access for early testers means the model's capabilities remain in limited circulation even after a national security directive, raising questions about the effectiveness of the halt order.
Reporting from 1 sources: GIGAZINE.
Despite the US government ordering Anthropic to halt distribution of Claude Mythos 5 and Claude Fable 5 on June 13, 2026, Bloomberg reports that some organizations that participated in early testing of Claude Mythos Preview still have access to the model. The early access was part of Project Glasswing, an initiative to use AI for cyber defense.
Anthropic released Claude Mythos Preview on April 7, 2026, to select organizations for cyber defense testing under Project Glasswing. On June 9, the company updated the model as Claude Mythos 5 and Claude Fable 5. Four days later, US national security authorities ordered Anthropic to stop providing both models to all foreigners, including its own foreign employees, citing suspicions of Chinese government-linked access and concerns raised by Amazon CEO Andy Jassy. Anthropic complied by halting distribution to all customers.
Bloomberg reported on June 18 that some early testers of Claude Mythos Preview still retain access to the preview version after the order. Project Glasswing, which grants multiple organizations access to high-risk AI models to identify vulnerabilities, continues to operate. The initial participants included AWS and other unnamed entities.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.