Anthropic Prepares Mythos 1 for Public Release via Claude Code and Claude Security

Summary

Anthropic is preparing to significantly broaden access to its Mythos security-focused model, marking a major shift from earlier guidance that suggested the model would remain restricted to limited use. The company has already deployed Mythos through Project Glasswing, its collaborative cybersecurity initiative launched last month, which has identified over 10,000 high or critical vulnerabilities in essential software. Now, signals point to Mythos 1 being prepared for preview release through Claude Code and Claude Security products, with a new dashboard for the latter surfacing vulnerabilities and triage results.

The broader availability represents both a strategic move and a confidence signal about the model's maturity for security applications. While Anthropic has indicated that Mythos-class models will require "far stronger safeguards" before general public release, the expanded rollout through enterprise and developer products signals readiness for real-world deployment. In parallel, Anthropic is preparing Claude Opus 4.8 for release, with select partners already conducting internal evaluations.

• Claude Opus 4.8 expected in the coming weeks, following April's Opus 4.7 release

Editorial Opinion

Anthropic's shift toward broader access for Mythos represents a calculated bet that the model is ready for real-world security work. By pairing security capabilities with developer-friendly tools like Claude Code and enterprise-focused Claude Security, Anthropic is making vulnerability discovery tangible rather than theoretical. The scale of vulnerabilities discovered through Project Glasswing—over 10,000 in the first month—suggests the model's value is already proving itself, a necessary precondition before enterprises (and eventually the public) will trust AI with critical security work.