0xensec Daily Roundup — May 25, 2026
Artificial intelligence continues to drive profound change in cybersecurity, as demonstrated by Anthropic’s Project Glasswing. In just a single month since launch, Anthropic’s collaborative AI-driven initiative—built around the Claude Mythos Preview model and supported by industry giants including AWS, Microsoft, Google, and the Linux Foundation—unearthed over 10,000 high-severity vulnerability candidates across more than 1,000 open-source projects. After human review, over 1,700 were deemed exploitable, with more than 1,000 confirmed as high- or critical-severity issues. These numbers, while a testament to Glasswing’s technical prowess, expose a growing and uncomfortable reality: the capacity to find flaws now vastly exceeds the industry’s collective ability to patch them [1].
Read more →