The reversal comes after over 100 cybersecurity experts petitioned the Trump administration to rescind its export control directive on Anthropic's frontier AI models.
Anthropic is getting the green light to deploy its Claude Mythos 5 AI model to over 100 institutions, marking a reversal of the US government’s export control directive that had effectively shut down access to the company’s most advanced AI systems just days after their launch.
The whiplash here is remarkable. Anthropic launched Claude Fable 5 and Claude Mythos 5 on June 9, 2026. Three days later, the US government told them to pull the plug on both models for everyone, including American employees. Now, the restrictions are being lifted.
How we got here #
The original export control directive, issued on June 12, forced Anthropic to disable access to both Claude Fable 5 and Claude Mythos 5 globally. The order applied not just to foreign nationals, which is typical for export controls, but to all users across the board.
Anthropic publicly stated that the directive stemmed from what it described as a misunderstanding regarding security vulnerabilities.
The backstory makes this more interesting. Anthropic had been cooperating with national agencies on vulnerability assessments in governmental systems. That collaboration apparently triggered concerns that led to the broad export control order.
Mythos 5 was specifically designed for cybersecurity and biological research applications within vetted institutions.
During the suspension period, Anthropic found a workaround for its Claude Opus 4.8 model, which wasn’t covered by the export control directive.
The pressure campaign that worked #
Over 100 cybersecurity experts petitioned the Trump administration to rescind the restrictive export directive. The lifting of restrictions allows Anthropic to deploy Mythos 5 to over 100 institutions. These are vetted organizations working in cybersecurity and biological research, not a wide consumer rollout.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our