Group Gains Access to Anthropic's Mythos AI Model on Launch Day via Vendor, No System Impact Reported

A small group of unauthorized users accessed Anthropic's Mythos AI model through a third-party vendor, Bloomberg reported. The group gained access to Mythos, an AI product designed for enterprise security, on the day it was publicly announced. Members made an educated guess about the model's online location based on knowledge of Anthropic's format for other models, according to Bloomberg.

Anthropic stated there is no evidence that the unauthorized access affected its systems. The company described Mythos as powerful enough to enable sophisticated cyberattacks. A private online forum gained access to Anthropic’s Mythos cybersecurity tool via the third-party vendor, Bloomberg reported.

Anthropic is investigating a report claiming unauthorized access to Mythos through one of our third-party vendor environments, an Anthropic spokesperson stated. The unauthorized group has been using Mythos regularly since gaining access. They provided evidence to Bloomberg in the form of screenshots and a live demonstration of Mythos.

The unauthorized group used access enjoyed by a person interviewed by Bloomberg. That interviewed person is employed at a third-party contractor that works for Anthropic, according to Bloomberg. Members of the unauthorized group are part of a Discord channel that seeks information about unreleased AI models, Bloomberg reported.

Anthropic found no evidence that the unauthorized activity impacted its systems. The group in question is interested in playing around with new models, not wreaking havoc with them, a source told Bloomberg. The unauthorized group tried a number of different strategies to gain access to the model.