Anthropic is upgrading the scanner inside Claude Security to Claude Mythos 5, and it's now in public beta for every Claude Enterprise customer. Hook up a GitHub repo, and Mythos 5 will scan for code flaws, returning CWE classifications, confidence scores, severity ratings, and fix suggestions.
Claude Security itself isn't new — when it hit public beta back in late April, it was powered by Claude Opus 4.7. What's different now is that Anthropic has slipped Mythos 5, a model that was previously locked down to tightly vetted users, into the same security product. That's no small deal: Mythos 5 doesn't just detect vulnerabilities, it can turn them into working attacks. Hence the access restrictions.
Enterprises still can't call Mythos 5 directly, though. The model only runs in the scanning backend; what you get is the report and recommended patches. If you then fix the code inside Claude Code, you're using whatever models your company already has, and any patch needs human approval. The scans bill tokens against your existing plan — no separate Mythos 5 purchase required.
Anthropic is also planning to wire Mythos 5 into partner security tools, and it's kicking off the Defender Advantage Fund, a $35 million pool of Claude credits to help open-source projects find and fix vulnerabilities.