The Mythos AI model, intended for enterprise-level cybersecurity, was illegally accessed by users through a private forum. The company Anthropic is investigating the incident.

The Mythos AI model, developed for enterprise-level cybersecurity applications, is part of a limited testing program intended only for selected companies. According to Bloomberg, a group of users managed to access the model through a third-party provider involved in the testing, facilitating access via a private online forum. Sources indicate that unauthorized users not only breached the system but also used the model repeatedly. A spokesperson for Anthropic confirmed that the company is analyzing the situation, but there is no evidence that the internal infrastructure has been compromised. The model, which has not been publicly launched due to potential risks, is being tested under Project Glasswing, involving partners such as Amazon, Apple, and JPMorgan Chase. Mythos AI is used for identifying cybersecurity vulnerabilities and strengthening system security.