Anthropic AI: Glasswing Project Boosts Cybersecurity

Anthropic's Project Glasswing is bringing advanced AI into vulnerability discovery on a large scale. This initiative uses the restricted-access Claude Mythos Preview model. Twelve launch partners and over 40 other organizations are participating. What's interesting is that the security perimeter is now expanding beyond traditional networks and endpoints. It includes code, AI agents, and the software supply chain. Anthropic has committed $100 million in usage credits to this effort. Palo Alto Networks, a launch partner, reported a significant increase in vulnerability disclosures. In one month, they disclosed 26 CVEs, covering 75 issues. This is a big jump from their usual rate of fewer than five CVEs per month. They estimate organizations have a three-to-five-month window to get ahead of adversaries using similar AI capabilities. CrowdStrike, another founding member, is focused on protecting AI agents themselves. Their 2026 Global Threat Report indicated an 89% year-over-year increase in attacks by adversaries using AI. The bottom line is that AI is rapidly changing how cybersecurity operates, creating both new tools for defense and new areas to protect.