Unauthorized Discord group gained access to Anthropic's Mythos model
12 hours ago
- #Cybersecurity
- #AI Models
- #Data Breach
- Unauthorized users have gained access to Anthropic's cybersecurity AI tool, Mythos, through a third-party vendor environment.
- Anthropic is investigating the incident but states no evidence suggests their own systems were impacted by the unauthorized access.
- The group, part of a Discord channel focused on unreleased AI models, accessed Mythos using knowledge of Anthropic's model naming conventions and third-party contractor credentials.
- Mythos was released under limited preview via Project Glasswing to select vendors like Apple to prevent misuse by bad actors, as it could be weaponized against corporate security.
- If confirmed, this breach could undermine Anthropic's efforts to ensure enterprise security through controlled tool distribution, posing reputational and operational risks.