San Francisco (United States) (AFP) - The American artificial intelligence company Anthropic announced on Tuesday that it is investigating unauthorized access to its powerful Mythos model, which the company fears could become a valuable tool for hackers.

First modification: 04/22/2026 - 11:11

Anthropic said weeks ago that it initially limited the release of Mythos to 40 large technology firms to give them a head start in patching cybersecurity vulnerabilities before they could be exploited by attackers.

According to the financial agency Bloomberg, a small group of users in a private online forum gained access to the model through the computer system reserved for Anthropic's external suppliers.

"We are investigating a report that unauthorized access to Claude Mythos Preview occurred through one of our third-party vendor environments," an Anthropic spokesperson told AFP.

Users gained access to Mythos through various means, including using permissions that one of the participants had as an employee of an Anthropic contracting company, Bloomberg reported.

Anthropic works with a small number of third-party vendors who assist in model development.

The company postponed the general release of Mythos, which it says can detect security gaps that have existed for decades in peer-vetted systems and automated tools.

It initially shared Mythos with some key players in the US technology and financial sectors, such as Nvidia, Amazon and JP Morgan Chase, to allow them to improve their security infrastructure.

But the company has been accused of exaggerating the capabilities of a technology that is its main business and the center of fierce competition with rival OpenAI.

© 2026 AFP

The content you requested does not exist or is no longer available.