AI "Mythos" Can Reportedly Find Flaws in Nearly Every Computer System

A new AI model from Anthropic, named "Mythos," has demonstrated a startling capability: it can automatically find critical security flaws in a vast range of the world's computer systems. The model scans and analyzes software, identifying vulnerabilities that could be exploited by hackers. Early internal tests suggest its success rate is exceptionally high, covering "almost every" system it examines. This breakthrough raises urgent questions for cybersecurity. While intended as a defensive tool for experts to fix weaknesses, the same technology could provide a roadmap for cybercriminals and hostile states. The core concern is "dual-use" risk. If details of such a powerful model were leaked, malicious actors could use it to find and attack flaws at an unprecedented scale and speed. Anthropic has not released Mythos publicly. The company emphasizes its focus on developing safeguards and strict access controls. The debate now centers on how such powerful defensive tools can be secured, ensuring they protect systems rather than expose them.