Some Unknown Group Is Reportedly Using Claude Mythos Without Permission
Gizmodo • • business
Key Points:
- Anthropic confirmed it is investigating unauthorized access to its unreleased AI model, Claude Mythos, reportedly through a third-party vendor environment.
- Bloomberg verified the breach by reviewing a live demo and screenshots provided by a member of the group responsible for the access.
- The group gained access by combining information from a data breach at AI startup Mercor with insider access from a contractor working with Anthropic.
- The unauthorized users, who operate via a Discord group using bots to scan GitHub for unreleased AI data, have been experimenting with Claude Mythos since April 7, coinciding with the announcement of Project Glasswing.
- According to the anonymous source, the group’s intent is exploratory rather than malicious, focusing on testing the model rather than causing harm.
