Anthropic's Claude Mythos Preview reportedly accessed by unauthorized users through third-party contractor
Anthropic’s new Claude Mythos Preview model appears to have been accessed by a small group of unauthorized users.
According to Bloomberg (April 21) and subsequent reporting from TechCrunch, Fortune, and Wired, the access was gained through a third-party contractor’s environment. One individual in the group reportedly had legitimate access via their employer (a vendor working with Anthropic) and, combined with educated guessing based on previously leaked information, the group was able to reach the model. They are said to have used it in a private Discord group.
Anthropic confirmed they are investigating the report but stated they have no evidence of access beyond the third-party vendor environment.
The model was being rolled out in a limited capacity through Anthropic’s Project Glasswing initiative to selected partners for defensive security research.
sources:
- Bloomberg: https://www.bloomberg.com/news/articles/2026-04-21/anthropic-s-mythos-model-is-being-accessed-by-unauthorized-users
- TechCrunch: https://techcrunch.com/2026/04/21/unauthorized-group-has-gained-access-to-anthropics-exclusive-cyber-tool-mythos-report-claims/
- Anthropic’s own assessment: https://red.anthropic.com/2026/mythos-preview/