Anthropic Launches Project Glasswing: $100M Cybersecurity Initiative Powered by Mythos AI Model

Anthropic has officially rolled out Project Glasswing, a major $100 million cybersecurity program that deploys its powerful (and previously restricted) Claude Mythos AI model to proactively identify software vulnerabilities before malicious actors can exploit them. The initiative is backed by major tech players including Google, Microsoft, and AWS.

Details: Mythos excels at finding zero-day vulnerabilities and defending critical infrastructure, but Anthropic had limited its public access due to fears it could be weaponized by hackers. Project Glasswing turns this capability into a defensive tool, with controlled access for partners to scan systems stealthily.

Impact: This marks a significant shift in AI’s role in cybersecurity — from potential threat to proactive shield — amid rising state-sponsored attacks (including recent Iran-linked incidents). It sparks a “cybersecurity reckoning” and highlights the dual-use dilemma of frontier AI. UK regulators are also rushing to assess risks from the Mythos model. For enterprises and governments, it accelerates adoption of AI-driven security but raises ethical and control questions about who gets access to such powerful tools.