UNAUTHORIZED USERS ACCESS ANTHROPIC'S MYTHOS MODEL

Anthropic is looking into claims that unauthorized users accessed its Mythos model, a powerful AI tool designed to identify cybersecurity vulnerabilities. The group obtained access through a third-party contractor portal and used internet research techniques to gain entry, Bloomberg reported. The unauthorized access began around the time Anthropic announced the Mythos model. The group involved reportedly has no malicious intent and is simply testing the model's capabilities, according to a person familiar with the matter. When contacted by TechCrunch, Anthropic confirmed it is investigating the claims but emphasized that it has found no evidence of system compromise. The company did not provide additional details about the scope of the unauthorized access or how long it persisted. Mythos represents a significant development in AI-powered cybersecurity tools. The model's ability to detect security flaws makes it a high-value target, which adds urgency to Anthropic's investigation. The breach highlights ongoing security challenges for AI companies managing access to powerful models. Third-party contractor portals, while necessary for business operations, create potential vulnerability points that attackers and curious users can exploit. Anthropirc has not disclosed whether the unauthorized users remain able to access the model or if access has been revoked. The company also has not indicated whether it plans to enhance security protocols for its contractor portal or other systems. This incident comes as major AI companies face increasing scrutiny over model security and unauthorized access. Anthropic's response and findings from its investigation could influence how other organizations approach access control for sensitive AI systems.