Published on June 23, 2026 at 08:44 Comment 6 𝕏 Print © Anthropic Claude Mythos, Anthropic’s AI deemed too powerful to be made public, has just exposed a 29-year-old security flaw. This vulnerability allowed an attacker to steal the credentials or passwords of other users sharing the same network. Present since 1997, it went unnoticed by human audits and traditional security tools. Claude Mythos, the most powerful version of Anthropic’s AI, continues to ac