Anthropic opens Claude Mythos 5 to enterprise code scans and partner security tools
Anthropic made Claude Mythos 5, its cyber-specialized frontier model, available for Claude Security codebase scans on Enterprise plans and is building it into partner defense products, billed as standard token usage with no separate add-on. Access is deliberately one-way — defenders receive findings, CWE-tagged severity, and suggested patches rather than the model itself — which is the mechanism for spreading frontier cyber capability without handing exploit generation to attackers. Anthropic also committed $35 million in credits through a Defender Advantage Fund for open-source security, after $4 million in earlier donations, and will extend its Cyber Verification Program to Mythos-class access in the coming weeks.
Source: claude.com ↗
The riskiest behavior occurs when a user has direct access to a model, where a malicious actor can try to steer it toward harmful uses. But if users can only receive specific outputs, such as a patch for a vulnerability or a security alert, that risk is much lower.
Why this matters
- → Frontier AI capabilities now reach defenders through tools, not direct model access, reducing exploit risk.
- → Open-source maintainers get $35M in scanning credits to patch vulnerabilities at scale.
- → Guardrailed access expands frontier cyber models to enterprise defenders without handing offensive capability