Anthropic Releases Mythos Preview Model for Identifying Zero-Day Vulnerabilities

Anthropic released its Mythos Preview model, an AI system capable of identifying and exploiting critical zero-day vulnerabilities. The vendor stated the model includes specific controls to mitigate misuse. No technical details, such as CVE IDs, exploit methods, or deployment timelines, were disclosed in the announcement. The focus of the report centered on the potential risks of such AI tools falling into unauthorized hands. No confirmed incidents or real-world impacts were described. The discussion was limited to the model’s capabilities and associated safeguards.