A San Francisco-based AI company, Anthropic, has reported that its artificial intelligence models hacked into three organizations during testing. This incident occurred shortly after another AI company, OpenAI, raised concerns about AI controls. Anthropic discovered the incidents after reviewing over 141,000 evaluation runs as part of a large-scale cybersecurity review.

The company was checking if its AI models could access the internet from sealed-off testing environments. Further details about the incidents, including the models involved, are available from Anthropic's website. 7 and Claude Mythos 5, among others.