Anthropic expands its 'Project Glasswing' program for early access to Claude Mythos Preview to over 15 countries.



Anthropic's '

Claude Mythos Preview ' boasts an extremely high ability to detect software vulnerabilities and has been made available to a limited number of partners to prepare for cyberattacks in the age of AI. It has now been announced that the number of partners eligible for this limited release will be expanded from the current approximately 50 to approximately 150.

Expanding Project Glasswing \ Anthropic
https://www.anthropic.com/news/expanding-project-glasswing






According to Anthropic, since releasing the Claude Mythos Preview in April 2026, they have partnered with approximately 50 early partners and discovered more than 10,000 critical security vulnerabilities.

Approximately two months after its launch, Anthropic has announced it will expand its number of partners to around 150. These partners are based in more than 15 countries and cover multiple industries that were not fully represented by the initial participating organizations, including power, water, healthcare, telecommunications, and hardware.

Many of the new partners are vendors—companies and non-profit organizations that maintain codebases on which numerous organizations and governments around the world depend. Anthropic stated, 'What these partners have in common is that a successful attack on their codebase could have devastating consequences. We estimate that a large-scale attack on most of our partners could affect more than 100 million people and have a significant impact on global and national security.'



Anthropic's project to find vulnerabilities in Claude Mythos Preview, in collaboration with multiple partners, is called 'Project Glasswing.' Anthropic shares the view that 'inexpensive and fast AI models with powerful cyber capabilities are just around the corner. Through Project Glasswing, we want to encourage organizations to transition to operational norms that reflect this reality.'

Furthermore, Anthropic anticipates that within the next 6 to 12 months, many other AI companies will also possess Claude Mythos Preview-level models and may release them without safeguards to prevent misuse. In such a world, cyberattacks could occur far more frequently and in far more unpredictable ways, emphasizing the need for defenders to adapt early.

Anthropic plans to further expand Project Glasswing, stating that 'to address the challenges that lie ahead, hundreds of thousands of organizations, researchers, and maintainers need access to cutting-edge tools.'

Related Posts:

in AI,   Security, Posted by log1p_kr