Anthropic has claimed that its high-risk new model Mythos experienced unauthorized access on its first day of testing.

According to Dynamic Beating monitoring, Bloomberg reported, citing sources familiar with the matter, that a small group of users accessed Anthropic's new AI model, Mythos, through a private online forum. Anthropic had previously claimed that Mythos was powerful enough to be used to launch dangerous cyber attacks.

On the same day Anthropic first announced its limited test of Mythos for a few enterprises, these users had already gained access. Since then, they have been regularly using the model, but not for cyber attacks. Sources provided screenshots to Bloomberg and gave an on-site demonstration of the model.