Outsiders breached Anthropic's Claude Mythos same day 'potentially dangerous' AI model was revealed: report - New York Post

A report indicates that unauthorized users gained access to Anthropic's Claude Mythos AI model on the same day it was publicly revealed. The breach occurred through methods including guessing the model's online address based on Anthropic's previous naming conventions for model releases. Anthropic has not yet issued a public statement about the incident.

The New York Post report states that a handful of individuals managed to breach the system using various tactics. Claude Mythos is Anthropic's cybersecurity-focused model, which the company had recently promoted as a tool for identifying vulnerabilities. The breach reportedly happened shortly after the model's unveiling to select partners and security researchers.

This incident follows recent public criticism from OpenAI CEO Sam Altman, who characterized Anthropic's marketing of Mythos as 'fear-based.' The breach raises questions about access controls for AI models marketed for security applications. No specific details about what the unauthorized users accessed or any potential data exposure have been disclosed in the initial report.