Rogue Group Gains Access to Anthropic’s Dangerous New Mythos AI

Bloomberg reported that a small private Discord group accessed Anthropic’s unreleased Claude Mythos preview through a third-party vendor environment.

Anthropic says it is investigating the incident and has not confirmed that a breach occurred.

The report raises concerns that a highly restricted AI model may have been exposed to unvetted users earlier than intended.

The incident highlights broader security risks around protecting advanced AI systems before full safeguards are in place.