Anthropic is adding Claude Mythos 5 to Claude Security, making the security-focused model available in public beta to all Claude Enterprise customers. The company describes Mythos 5 as its “most capable security model,” although the announcement does not provide independent testing to support that characterization.
Claude Security can scan GitHub repositories for vulnerabilities across files, while analyzing how components interact. Anthropic claims each finding includes a CWE category, confidence and severity ratings, along with a suggested fix.
Suggested patches open in Claude Code on the web, using the models already available to the team. Mythos 5 operates behind the scan rather than being exposed for direct use, with the service returning findings instead. Anthropic also states that scans are billed as standard token usage under an existing plan.
The scan interface allows configuration of a repository, branch, scan scope, model and effort level. Available controls include adding a directory, canceling a scan and starting it; the interface lists Mythos 5 with a Standard effort setting.
Access remains limited to Enterprise customers, a restriction that drew criticism in replies to the announcement. Some users questioned why Anthropic continues to limit direct access to its security-related models while offering their capabilities through an enterprise service.
A Japanese-language post responding to the launch connected the feature to a Anthropic states that it is working with partners to integrate Mythos 5 into security products and services. The company also mentions a $35 million Defender Advantage Fund for open-source security and plans to expand its Cyber Verification Program.
Source: Claude on X
