Anthropic opens Mythos 5 to defenders, adds $35M OSS security fund
The frontier cyber model can now scan enterprise codebases for tagged bugs and patches; the fund and verification-program access are the door for solo builders.
Copy markdown
Mythos 5 now scans your codebase from inside Claude Security
Enterprise teams can point Claude's frontier cyber model at a repo and get back CWE-tagged findings, severity and confidence scores, and suggested patches — billed as ordinary token usage (reportedly ~$10 in / $50 out per million). It's an admin-enabled public beta: you get alerts and patches, never raw model access.
$35M in Claude credits for open-source security
The new Defender Advantage Fund puts $35M in API credits — not cash — behind teams that find and patch vulnerabilities in open-source software or automate the scan-and-patch loop. Anthropic is starting with a few larger pilot grants and says application details land in the coming weeks, so if you maintain OSS security tooling, get on the list.
The Cyber Verification Program is opening up on Opus and Sonnet
Vetted defenders can apply now for reduced safeguards on Opus and Sonnet, with broader dual-use capabilities and eventual Mythos-class access on the roadmap. This is the individual-builder door: get verified and you can build security tooling the default models would refuse.