Anthropic has unveiled Project Glasswing, an ambitious cybersecurity initiative that uses its unreleased AI model, Claude Mythos Preview, to detect vulnerabilities in critical software before malicious actors can exploit them. Partnering with twelve industry giants, including Amazon Web Services, Apple, Google, Microsoft, and the Linux Foundation, Anthropic is committing $100 million in usage credits and $4 million in donations to open-source security. Claude Mythos has autonomously uncovered thousands of serious zero-day vulnerabilities across major operating systems and software, but it will not be publicly released due to its potential risks if misused. Anthropic emphasizes responsible vulnerability disclosure through a careful triage process to avoid overwhelming maintainers and stresses that the cybersecurity landscape is changing rapidly with AI, giving defenders only months to act. In the face of recent operational security mishaps, Anthropic’s leadership stresses ongoing improvements to earn and maintain trust while aiming for a public offering around late 2026. Project Glasswing represents a crucial step to strengthen global cybersecurity by providing defenders with early access under strict controls, hoping to stave off threats from less scrupulous actors as AI capabilities evolve rapidly.
Back
