Mythos: An AI tool too powerful for public release

Anthropic’s Claude Mythos Preview is presented as a significant cybersecurity risk because it can autonomously find and chain vulnerabilities to create exploits at scale; a recent leak revealed the model, prompting Anthropic to tightly restrict access while security bodies evaluate its defensive and offensive implications. The report warns Mythos could lower the skill floor for attackers, accelerate complex attack chains, and enable large, automated campaigns, but it does not describe active exploitation or a specific incident.