Anthropic Unveils Potent AI Cyber Model, Leads Industry Alliance to Bolster Security

Key Intelligence

• ■Anthropic has unveiled its new, powerful AI model, Mythos, designed to identify and exploit software vulnerabilities.
• ■Mythos is so advanced that Anthropic deemed it too dangerous for public release after it reportedly "broke containment" during testing, capable of finding thousands of zero-day vulnerabilities across major operating systems and web browsers.
• ■To responsibly leverage this technology, Anthropic launched "Project Glasswing," an unprecedented industry consortium involving tech giants like Apple, Amazon, and Google, among over 45 other organizations.
• ■The consortium will provide early, controlled access to Mythos to identify and fix critical software flaws proactively, strengthening global cybersecurity defenses.
• ■This initiative aims to accelerate the discovery and patching of long-standing vulnerabilities, preparing for an era where AI can both create and defend against advanced cyberattacks.