Anthropic’s restricted Claude Mythos model may be coming to Claude Code

Key Points:

• Anthropic is preparing to publicly roll out "Mythos," a new AI model with advanced capabilities in computer security, originally announced in April as a restricted model due to significant security risks.
• Mythos demonstrates major improvements in code reasoning and autonomy, enabling it to automatically develop highly professional cyberattacks, which poses a severe risk to global digital infrastructure.
• To mitigate potential exploitation of unpatched vulnerabilities in popular software, Anthropic delayed Mythos's public release until a robust guardrail system was developed, with evidence of Mythos integration appearing briefly in public versions of Claude Code and Claude Security.
• Anthropic is collaborating with other companies through a project called "Glasswing," using the unreleased Mythos model to identify AI-driven exploits and has helped around 50 partners uncover over 10,000 high- or critical-severity vulnerabilities within its first month.
• The company currently offers several AI models including Claude Opus 4.7, Opus 4.6, Opus 4.5, Sonnet 4.6, and Haiku 5.5, while preparing Mythos for a controlled public rollout.