"Anthropic says one of its AI models has found thousands of high-severity zero-day vulnerabilities across every major operating system and web browser. The model is called Claude Mythos, and it is the centerpiece of a new Anthropic cybersecurity initiative called Project Glasswing."