Anthropic Claude Security Now Uses Mythos 5 for Enterprise

Anthropic has integrated its most advanced cybersecurity model, Claude Mythos 5, into its Claude Security product, making it available to enterprise customers as of August 21, 2026. Previously accessible only to vetted security professionals through Project Glasswing, Claude Mythos 5 will now power vulnerability scanning for organizations using Claude Enterprise. This new iteration of Claude Security connects directly to GitHub repositories, analyzes data flow across files, and provides detailed findings. Each finding includes a Common Weakness Enumeration (CWE) category, confidence and severity ratings, and actionable suggestions for patches. A key design feature of Claude Security is that it returns scan results rather than engaging in conversational prompts, thereby preventing the model that identifies vulnerabilities from being used to generate malicious code or exploits. The service is currently in public beta exclusively for Claude Enterprise customers, with no additional charges beyond standard token usage under their existing plans. Administrators can enable Claude Security through the admin console, and users can initiate scans via the claude.ai/security portal. This feature is specifically targeted at enterprise-tier organizations that have a Claude Enterprise agreement and host their code on GitHub. Smaller businesses, including startups and mid-market companies on Pro, Max, or Team plans, will not have access to the Mythos 5-powered scanning. Separately, a Claude Security plugin for Claude Code is also in beta for all Claude Code users, but it operates using the models available within the user's account, not the specialized Mythos 5 model. Anthropic has identified several key industries that stand to benefit from this enhanced security scanning, including hospitals, utility providers, financial systems, and the broader software supply chain. The company suggests that any regulated organization already employing Static Application Security Testing (SAST) in their Continuous Integration (CI) pipelines, such as those in healthcare, fintech, government contracting, and critical infrastructure, would find this tool particularly valuable. Potential applications for Claude Security include pre-release codebase audits, comprehensive reviews of legacy code, scheduled regression scanning to detect new vulnerabilities, efficient triage of existing vulnerability backlogs, and targeted directory-scoped scans on services deemed high-risk. This announcement coincides with three other updates from Anthropic, underscoring a broader push in AI security and developer tooling.