§ safety · storyline
Anthropic's Mythos Preview finds exploit chains earlier models missed
Anthropic's Mythos Preview security model identified exploit chains in Cloudflare's codebase that earlier frontier models missed, during a trial across more than 50 internal repositories.
Cloudflare tested Anthropic's security-focused AI model Mythos Preview across more than 50 of its own code repositories as part of Project Glasswing.
The article Cloudflare says Anthropic's Mythos Preview finds exploit chains that earlier frontier models missed appeared first on The Decoder.
§ sources1 publication · timeline below