Anthropic embeds Claude Mythos 5 in security tools, pledges $35M in credits

What's the deal? AnthropicDealroom has a profile for this one. Try Dealroom → is embedding its Mythos 5 model into cybersecurity products for security operations, incident response, threat intelligence, and detection engineering. It is also launching Claude Security in public beta for Claude Enterprise customers, with a $35 million credit fund for open-source projects.

How does it work? An administrator enables code scanning in the Claude admin console, after which users select a repository and receive findings classified by CWE, confidence, severity, and a recommended fix. Scans are billed as normal token usage under existing plans.

What's the guardrail? Rather than an open interface that could be steered toward exploit development, Mythos 5 returns task-specific results such as vulnerability findings or patch suggestions. Scans are limited to code the customer owns and do not grant the model access to other Claude surfaces.

Every patch requires human sign-off. Teams can open Claude Code on the web to implement proposed changes, but a person must review and approve each one before deployment.

Where does the $35M go? Anthropic's Defender Advantage Fund will provide Claude credits to organizations helping maintainers fix live vulnerabilities, automate scanning and patching, and build defenses against broader classes of attacks. The program starts with a small set of pilot grants, with the first recipients due in the coming weeks.

What's the endgame? Anthropic plans to expand its Cyber Verification Program, giving approved organizations broader dual-use capabilities on Claude Opus and Sonnet, including fewer interruptions during authorized security work, followed by Mythos-class access for defensive tasks. It is also continuing Project Glasswing with US government partners to bring Mythos access to critical-infrastructure organizations that meet strict security requirements.

The signal: By restricting its most capable model to defensive, human-supervised tasks, Anthropic is positioning Mythos 5 as a security tool rather than a general-purpose assistant. The credit fund and verification tiers suggest a strategy of vetting who gets access to dual-use capabilities as AI moves deeper into security workflows.

Read more: 4sysops.com

Image credit: Department for Science, Innovation & Technology