By Dan Milmo
Publication Date: 2026-04-22 08:58:00
AI developer Anthropic has confirmed it is investigating a report that unauthorized users gained access to its Mythos model, which it says poses cybersecurity risks.
The US startup made the statement after Bloomberg reported on Wednesday that a small group of people had accessed the model, which was not made available to the public due to its ability to enable cyberattacks.
“We are investigating a report involving unauthorized access to Claude Mythos Preview through one of our third-party environments,” Anthropic said.
Bloomberg said a “handful” of users had gained access to Mythos in a private online forum the same day Anthropic said it was being shared with a small number of companies, including Apple and Goldman Sachs, for testing purposes.
It was reported that the unnamed users came to Mythos through access that one of them had as a third-party employee of Anthropic and through the use of methods used by cybersecurity researchers.
The group has…