Anthropic Integrates Mythos 5 into Claude Security, Available for Enterprises but Not Directly Callable

Anthropic has upgraded the underlying scanning model of Claude Security to Claude Mythos 5 and opened public testing to all Claude Enterprise customers. After enterprises access the GitHub repository, Mythos 5 will scan for code vulnerabilities, returning CWE classifications, confidence levels, severity ratings, and remediation suggestions. Since its public testing began at the end of April, Claude Security initially used Claude Opus 4.7, and this upgrade integrates Mythos 5 into the existing security product. Mythos 5 can identify vulnerabilities and convert them into actionable attacks, which is why it was previously only available to vetted organizations. Enterprises still cannot directly call Mythos 5; the model runs only in the scanning backend, and users receive vulnerability reports and remediation suggestions. When modifying code in Claude Code later, enterprises must still use their existing models, and patches must be manually approved. Scanning is charged normally according to the existing package without the need to purchase Mythos 5 separately. Anthropic also plans to integrate Mythos 5 into partners' security products and launch a $35 million Claude Credit Defender Advantage Fund to help open-source projects find and fix vulnerabilities.