Anthropic expands Mythos

Anthropic is expanding Project Glasswing, a joint industry initiative designed to find and fix critical software vulnerabilities with AI, to around 150 new organizations across more than 15 countries. According to TechCrunch, the project is built around Claude Mythos, an advanced Anthropic model that the company describes as its most powerful yet for cybersecurity work. The model is designed to scan codebases and identify serious flaws, including zero-day vulnerabilities.

The expansion marks a significant step in the use of frontier AI for defensive cybersecurity. In early April, Anthropic gave 50 initial partners, including the U.S. government, access to Claude Mythos Preview. Those partners were able to use the model to examine their software for vulnerabilities and security weaknesses. The new phase broadens the program to organizations that operate or support critical infrastructure.

The sectors now included are especially sensitive: power, water, healthcare, communications and hardware. Anthropic said these areas were not well represented in the first cohort. Many of the new participants maintain codebases that other companies, governments and citizens rely on. That makes the project more than a technical experiment: it is also a test of how AI can be used to protect systems that society depends on every day.

Anthropic said the common factor among the partners is that a successful attack on their software could have catastrophic consequences. For many of them, the company estimates that a major cyberattack could affect more than 100 million people and carry serious implications for national and global security.

According to the Financial Times, the expanded group includes organizations in U.S.-aligned countries such as Australia, Canada, France, Germany, Italy, Switzerland, the Netherlands, Spain, Belgium, Sweden, India, Japan, New Zealand and South Korea. Reported participants include Okta, Samsung, SK Hynix, SK Telecom, NATO and the European Union Agency for Cybersecurity, ENISA. TechCrunch noted that it had asked Anthropic to confirm those details.

The move also reflects growing competition in AI-powered cybersecurity. Anthropic expects other AI companies to develop models with similar capabilities, which is why it says Project Glasswing is partly about establishing safeguards early. For governments and critical industries, the message is clear: AI is no longer only a productivity tool. It is becoming part of the security layer around essential infrastructure.