Anthropic Launches Project Glasswing to Secure Critical Software With Claude Mythos Preview Model.

Anthropic has launched Project Glasswing, an initiative to identify and remediate security vulnerabilities in the world's most critical software systems using a new frontier AI model called Claude Mythos Preview. The company announced on LinkedIn that the model has already uncovered thousands of high-severity vulnerabilities across major operating systems and web browsers.

The project brings together a consortium of technology and security industry leaders, including Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks. Anthropic is committing up to $100 million in Mythos Preview usage credits for partners and over 40 other organizations that maintain critical software infrastructure, including open-source projects.

Unlike previous model releases, Anthropic does not plan to make Claude Mythos Preview generally available. The company cited the dual-use nature of vulnerability detection capabilities, which could be exploited by malicious actors if widely distributed. Instead, Anthropic is focusing on deploying the model through controlled partnerships for defensive security work.

The company stated it will first develop safeguards that reliably block the model's most dangerous outputs before attempting broader deployment. These safeguards will be tested with an upcoming Claude Opus model, building on the recent Opus 4.7 release that shipped with enhanced vision capabilities and self-verification features.

Mythos Preview represents a significant capability leap in automated vulnerability detection. According to Anthropic's technical report, the model performs at a level comparable to highly skilled security researchers. The vulnerabilities discovered span every major operating system and web browser, highlighting both the breadth of existing security gaps and the potential for AI-assisted defensive work.

Anthropic emphasized that no single organization can address cybersecurity challenges alone, calling for coordination across industry, open source communities, academic researchers, and government entities. The company published a detailed technical report on the software vulnerabilities and exploits discovered by Mythos Preview, along with a system card documenting the model's capabilities and limitations.

The Project Glasswing announcement comes as AI capabilities in both offensive and defensive security applications advance rapidly. While Anthropic acknowledged that models with these capabilities will likely become more widespread, the company argued there are strong reasons for optimism about AI's role in defensive cybersecurity work.

Anthropic stated it will report publicly on findings from Project Glasswing as partner organizations use Mythos Preview to identify and remediate vulnerabilities in their systems. The initiative marks a shift toward more controlled deployment of frontier AI capabilities in security-sensitive domains, prioritizing coordination with established security organizations over broad public access.