Anthropic previews Mythos in cybersecurity push with Amazon, Apple, Microsoft and others
Anthropic opens limited preview of Mythos
Anthropic on Tuesday released a preview of Mythos, a new frontier AI model the company says will be used in a tightly controlled cybersecurity initiative called Project Glasswing. The program will bring together 12 partner organizations to use the model for defensive security work, including scanning first-party and open source software for vulnerabilities.
Anthropic described Mythos as one of its “most powerful” models in a previously leaked memo, and said it is a general-purpose model for its Claude AI systems with strong agentic coding and reasoning abilities. While the model was not specifically trained for cybersecurity, the company said it can help identify weaknesses in code and critical software systems.
Anthropic said Mythos found “thousands of zero-day vulnerabilities” over the past few weeks, including many it described as critical. The company added that many of the flaws appear to be one to two decades old.
Major tech companies join Project Glasswing
The partner organizations in the initiative include Amazon, Apple, Broadcom, Cisco, CrowdStrike, the Linux Foundation, Microsoft and Palo Alto Networks. Anthropic said those partners will eventually share what they learn from using Mythos so the broader tech industry can benefit from the findings.
The preview will not be generally available, though Anthropic said 40 organizations will get access to the Mythos preview outside the core partnership.
Anthropic also said it has had “ongoing discussions” with federal officials about Mythos. Those talks come as the company remains in a legal dispute with the Trump administration after the Pentagon labeled Anthropic a supply-chain risk over its refusal to allow autonomous targeting or surveillance of U.S. citizens.
News of Mythos had already surfaced in a security incident reported last month by Fortune, after a draft blog about the model — then referred to as Capybara — was found in an unsecured cache of documents on a publicly inspectable data lake.
The limited rollout marks Anthropic’s latest effort to position its most advanced models as tools for defensive security, rather than broader public deployment.
Sources:
Read more tech news on the Doppler VPN Blog.