A powerful new AI model developed by Anthropic called Mythos — feared to outperform humans at finding and exploiting software vulnerabilities — has experts worried that, once widely available, it could give malicious actors an unprecedented ability to punch through critical IT infrastructure. Its developers are currently restricting access to a limited group of technology firms and unnamed organizations while they assess the security risks.
AI allows hackers to strike faster and at greater scale than traditional methods — compressing what once took hours of manual work into seconds, and enabling more simultaneous attacks across more targets. Generative AI also enables highly convincing phishing emails, voice cloning and deepfake videos that can bypass security checks that rely on human intuition.
Cybersecurity specialists are also using AI to monitor networks and flag unusual behavior faster than any human analyst could.