Anthropic's Claude Mythos might get public release

Anthropic says its Claude Mythos cybersecurity model could eventually be made public, but only after stronger safeguards are in place.

The company plans to first expand access to U.S. and allied governments, while acknowledging current protections are not yet sufficient to prevent misuse.

Mythos has already shown strong vulnerability-finding ability, scanning open-source projects and uncovering tens of thousands of flaws.

One serious bug it found in wolfSSL has since been fixed.

A broader release could help defenders find bugs faster, but it also raises misuse risks for open-source maintainers and security teams.