Some early testers retained access to Anthropic’s Mythos Preview despite a US government order that led to the shutdown of other versions of the model.
Companies including Cisco and Dragos confirmed continued access, while some invited organisations such as Enisa were denied entry after the crackdown.
The US had ordered Anthropic to suspend access over national security concerns tied to the model’s ability to identify software vulnerabilities and potential jailbreak risks.
A handful of firms given early access to test Anthropic's Mythos AI model have kept that access, even after a US government order shut the system down everywhere else, Bloomberg News reported.
Anthropic was forced to disable its flagship Fable 5 and Mythos 5 models worldwide after receiving an export control directive from the US government. The company said it had roughly 90 minutes to comply. Within that window, some of the world's most advanced AI systems disappeared from public use almost without warning.
1 June 2026
Get the latest issue of Outlook Business
Dragos and Cisco Systems confirmed to Bloomberg News that they had retained access to Mythos Preview despite the broader block. It was not immediately clear how Anthropic was determining access for individual members of its Glasswing programme.
Notably, Enisa, the European Union's cybersecurity agency, had been invited to join Glasswing ahead of the US government's block. It was informed on Friday that it would no longer be given access, the report said.
Why the Shutdown Happened
Anthropic was forced to disable its flagship Fable 5 and Mythos 5 models worldwide after receiving an export control directive from the US government on June 12. The company said it had roughly 90 minutes to comply. Within that window, some of the world's most advanced AI systems disappeared from public use.
The directive ordered Anthropic to suspend access to Fable 5 and Mythos 5 for foreign nationals, citing national security concerns. According to the company, authorities believed there may be ways to bypass safeguards built into the models and use them to identify software vulnerabilities.
Anthropic had limited Mythos Preview to about 200 organisations, including the US government, under its Glasswing programme, after the model identified thousands of software vulnerabilities. A less powerful, publicly released version of Mythos was also disabled following Washington's directive to suspend access for all foreign nationals due to national security concerns.
Anthropic argued that the concerns behind the order stemmed from a narrow and disputed "jailbreak" scenario. However, the company said it could not immediately separate all affected users from its global customer base. It chose the only option it believed would ensure compliance, shutting down access for everyone.
The result was unprecedented in scale. Developers, enterprises, researchers and paying customers around the world lost access to the models almost instantly. Even some of Anthropic's own employees were affected, multiple reports said.
Anthropic's relationship with the US government had already strained this year, after the company refused to let the US military use its AI models for domestic surveillance and fully autonomous weapons systems. The company is currently working towards an IPO. Anthropic said access to all its other Claude models remained unaffected by the directive.
