Apr 14, 2026 Claude Mythos: The AI Model Too Dangerous to Release Anthropic built its most capable model ever — then decided not to release it publicly. Here's what Mythos can do, why it matters, and what Project Glasswing means for cybersecurity.