Anthropic's method for finding vulnerabilities involved writing a script that used an agent to trace the system from every file in the codebase, looking for CVEs and potential exploits.

Anthropic believes that providing software developers with advanced AI tools, such as Claude Mythos Preview, offers a collective headstart in identifying and remediating vulnerabilities more rapidly.

The recent increase in CVEs (Common Vulnerabilities and Exposures) is partly attributed to security researchers using LLMs to bridge knowledge gaps about various software systems.

Anthropic's testing found that Mythos Preview can identify and exploit zero-day vulnerabilities in every major operating system and web browser when directed by a user.

Anthropic's Mythos model has the ability to chain together three, four, or sometimes five vulnerabilities to create sophisticated exploits.

Anthropic's Mythos model found a 16-year-old bug in FFmpeg that was missed by automated tools after 5 million scans.