Anthropic's Mythos model accessed unauthorized

A security breach has allegedly exposed Anthropic's 'Mythos' AI model, a tool designed to identify and exploit vulnerabilities in major operating systems and browsers.

Bloomberg reports that an unknown group has been using the model regularly, though not yet for malicious cybersecurity purposes.

The incident has prompted concerns at the Reserve Bank of Australia and other institutions about the risks of frontier AI tools falling into the wrong hands.

Anthropic maintains that its internal systems remain secure but is working to determine how the access was obtained.