Reports indicate that despite the US government issuing a regulatory order against Anthropic's 'Claude Mythos,' members of Anthropic's 'Project Glasswing' have still been provided with a preview version.

Anthropic released ' Claude Mythos 5 ' and ' Claude Fable 5, ' updated versions of its high-performance AI model ' Claude Mythos Preview, ' which was capable of sophisticated cyberattacks, on June 9, 2026. However, following instructions from the US government's national security authorities, Anthropic announced the discontinuation of both models on June 13, 2026 (Japan time). However, Bloomberg has revealed that some organizations that participated in the initial testing of Claude Mythos Preview still retain access to the models.

Early Users of Anthropic Mythos Still Have Access After US Order - Bloomberg
https://www.bloomberg.com/news/articles/2026-06-19/early-users-of-anthropic-mythos-still-have-access-after-us-order

Early users of Anthropic's Mythos still have access after US order: Report - The Economic Times
https://economictimes.indiatimes.com/tech/artificial-intelligence/early-users-of-anthropics-mythos-still-have-access-after-us-order-report/articleshow/131846387.cms

Early users of Anthropic's Mythos still have access after US order, Bloomberg News reports | Reuters
https://www.reuters.com/business/early-users-anthropics-mythos-still-have-access-after-us-order-bloomberg-news-2026-06-19/

On April 7, 2026, Anthropic announced 'Claude Mythos Preview,' a high-performance AI model capable of sophisticated cyberattacks, and made it available to select organizations with the aim of strengthening their cyber defense capabilities. Subsequently, on June 9, 2026, Claude Fable 5 and Claude Mythos 5 were released as updated versions of Claude Mythos Preview. Claude Mythos 5, provided only to select organizations, is positioned as a limited version with no security restrictions, while the publicly released Claude Fable 5 is positioned as a retail version with security restrictions.

The official version of 'Claude Mythos' has finally been released, and 'Claude Fable,' a version with no user restrictions, has also been released, making it available to everyone - GIGAZINE

However, Anthropic announced the discontinuation of Claude Fable 5 and Claude Mythos 5 at approximately 9:50 AM JST on June 13, 2026. According to Anthropic's announcement, the US government's National Security Agency issued an export control order ordering Anthropic to cease providing Claude Fable 5 and Claude Mythos 5 to all foreigners, including Anthropic's foreign employees, both inside and outside the United States. In order to comply with the order, Anthropic decided to discontinue providing Claude Fable 5 and Claude Mythos 5 to all customers. The US government's order to cease operations has been reported to have been triggered by concerns that the models may have been accessed by groups linked to the Chinese government, and by concerns expressed by Amazon CEO Andy Jassy to government officials that the models could output information that could be misused for cyberattacks.

Anthropic has stopped distributing 'Claude Fable 5' and 'Claude Mythos 5' at the direction of the US government - GIGAZINE

Meanwhile, Bloomberg reported on June 18, 2026, that some companies to which Anthropic provided models for early testing of Claude Mythos Preview still retain access to Claude Mythos Preview even after regulatory orders were issued for Claude Fable 5 and Claude Mythos 5.

When Anthropic announced the Claude Mythos Preview, it pointed out that 'given the speed of AI development, it's only a matter of time before AI with a high risk of exploitation becomes widely available,' and launched ' Project Glasswing ' as an urgent effort to use AI for defensive purposes before attackers begin to exploit it. Project Glasswing is an initiative to grant multiple organizations access to AI models such as Claude Mythos Preview to help them discover vulnerabilities. Early participating organizations included AWS, Anthropic, Apple, Broadcom, Cisco, CrowdStrike, Google, JP Morgan Chase, Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks, and the number of participating organizations has since expanded to around 200.

Cisco and Dragos, a security company for industrial control systems, have confirmed to Bloomberg that they 'still maintain access to Claude Mythos Preview.'

Meanwhile, the European Union Cybersecurity Authority (ENISA), which had been invited to participate in Project Glasswing, stated that it received notification on June 19, 2026, that it would no longer be granted access. ENISA commented that it is continuing discussions with Anthropic while also considering other AI companies.

Anthropic has not responded to media requests for comment regarding its access management methods for Project Glasswing participating organizations.