NATO, EU security agencies and organizations from more than 15 U.S.-friendly countries will gain access to Anthropic’s Claude Mythos artificial intelligence as part of a major expansion of Project Glasswing, a security initiative designed to detect critical vulnerabilities in software before hackers can exploit them.
NATO, EU security agencies and organizations from more than 15 U.S.-friendly countries will gain access to Anthropic’s Claude Mythos artificial intelligence as part of a major expansion of Project Glasswing, a security initiative designed to detect critical vulnerabilities in software before hackers can exploit them.
This was reported by TechCrunch and journalists from The Financial Times, who, relying on their own sources, also clarified the list of new participants, among whom were key allies in the field of global security.
Project Glasswing is powered by Claude Mythos, Anthropic’s most powerful AI model to date. According to the developers, it can detect thousands of zero-day vulnerabilities within a few weeks. In early April 2026, Anthropic made a pre-release version of Mythos available to the first 50 partners, including the US government, to scan their codebases.
Now, the project is expanding to include about 150 new organizations. The new wave includes critical sectors not previously represented: energy, water, healthcare, communications, and hardware manufacturing. The tool will be available to companies and nonprofits whose code other businesses and governments rely on.
«What each partner has in common is that a successful attack on their codebase could have catastrophic consequences,» Anthropic said. The company estimates that a large-scale disruption to these systems could affect more than 100 million people, posing a direct threat to both national and global security.
According to media reports, the geography of the project includes organizations from Australia, Canada, France, Germany, Italy, Switzerland, the Netherlands, Spain, Belgium, Sweden, India, Japan, New Zealand and South Korea. Among the specific participants, sources name the American identity management platform Okta, South Korean giants Samsung, SK Hynix and SK Telecom, as well as the EU cybersecurity agency ENISA and the NATO military alliance.
Anthropic is rushing to deploy security measures within Project Glasswing, as it expects competitors to soon create models with similar power. Immediately after Claude Mythos’ presentation, the company’s main competitor, OpenAI, released its own GPT-5.5-Cyber model, focused on cybersecurity tasks, and handed it over to a large group of partners for testing.
It was previously reported that Anthropic’s specialized Claude Mythos artificial intelligence model helped a team of researchers uncover a critical vulnerability in Apple systems. The version of the AI, which the company currently provides only to select partners, allowed specialists to bypass macOS kernel memory protection on the latest M5 processors.
