Report claims unauthorized group accessed Anthropic’s proprietary cyber tool Mythos

A group of unauthorized users reportedly gained access to Mythos, Anthropic’s recently announced cybersecurity tool.

Much has been said about Mythos and its powers. Mythos is an AI product designed for enterprise security, and the company says it can be a powerful hacking tool in the wrong hands. Bloomberg now reported that a “private online forum” whose members are not publicly disclosed obtained access to the tool through a third-party vendor.

An Anthropic spokesperson told TechCrunch that the company is “investigating reports of unauthorized access to Claude Mythos Preview through one of our third-party vendor environments.” The company said it has so far found no evidence that the alleged fraud affected Anthropic’s systems in any way.

The unauthorized group tried various strategies to gain access to the model, including taking advantage of the “access” enjoyed by the person interviewed by Bloomberg. The person is currently employed by a third-party contractor working for Anthropic, the report said.

The group’s members are part of a Discord channel seeking information about unreleased AI models, the media reported. The group has been using Mythos regularly since accessing it and provided evidence to Bloomberg in the form of screenshots and live demonstrations of the software.

The group allegedly accessed the tool the same day as the official announcement and “made educated guesses about the models’ online locations based on their knowledge of the formats Anthropic has used for other models,” Bloomberg reported. The group in question is “interested in tinkering with the new model rather than wreaking havoc with it,” a source told the outlet.

Mythos was released to select vendors, including major companies like Apple, as part of an initiative called Project Glasswing. The limited release of this model was designed to prevent its use by malicious parties. Anthropic said the tool could be used as a weapon rather than to strengthen a company’s security.

If true, Mythos abuse could pose a problem for Anthropic, which provided an exclusive release to allay the company’s concerns about enterprise security.