On June 2, 2026, Anthropic announced it was expanding Project Glasswing, its defensive cybersecurity initiative, to roughly 150 new organizations across more than 15 countries. Project Glasswing uses Claude Mythos Preview, an unreleased frontier model, to identify security vulnerabilities in critical software so they can be fixed before attackers exploit them. Anthropic said the initial set of around 50 partners had already found more than 10,000 high- or critical-severity security flaws.
The expansion targets organizations whose codebases underpin essential infrastructure, where Anthropic estimates a single successful attack could affect more than 100 million people. The effort reflects a broader argument the company has been making in 2026: that AI has reached a level of coding capability where it can surpass most humans at finding and exploiting software vulnerabilities, which makes deploying that capability defensively a priority rather than an option.
For business and general readers, the significance is that frontier AI is now a practical cybersecurity tool at industrial scale, not a demo. The same capability that could let attackers discover zero-days at machine speed is being turned toward hardening the software that critical infrastructure depends on, reshaping how large organizations think about software security.