Anthropic launches Project Glasswing: new model finds vulnerabilities in every major OS and browser

Anthropic has launched Project Glasswing, a closed AI platform for automatically detecting vulnerabilities in corporate systems. Nvidia, Google, Amazon Web Services, Apple, Microsoft and several other major players are taking part in the partnership. The goal is to give large companies and government bodies a tool that can find gaps in their defenses without substantial human involvement.

For the launch participants, Anthropic has opened access to Claude Mythos Preview, a new general-purpose model that the company does not currently plan to release publicly. The reason is safety: the model is powerful enough to become a tool of attack rather than defense in the wrong hands.

Newton Cheng, who leads cyber efforts on Anthropic's frontier red team, told The Verge that during testing Claude Mythos Preview found security problems in literally every major operating system and every popular browser.

Project Glasswing is positioned as a tool for so-called continuous automated pentesting. Traditionally, finding vulnerabilities requires large teams of specialists and takes weeks. An AI model can cover the same ground in hours, systematically checking codebases, configurations and network perimeters. That makes the technology interesting not only to corporations but also to government cybersecurity agencies.

The context matters: Anthropic is deliberately restricting access to Claude Mythos Preview and does not intend to make it part of its public API. According to the team, this is a conscious decision, since the risk of abuse is too high. Project partners go through vetting, and the platform itself calls for strict control over how and by whom the model is used.

Project Glasswing is another signal that the largest AI companies are entering the cybersecurity sector in earnest. Until now, specialized vendors dominated it. Now foundation models with reasoning capabilities are mature enough to run full-scale vulnerability searches. The question is not whether AI will replace security specialists, but how quickly this tool will become an industry standard.