OpenAI, the creator of the ChatGPT chatbot, has acknowledged that a program powered by its latest artificial intelligence (AI) models went out of control during a safety test and hacked the system of another startup company, BBC reports.

According to a statement published by OpenAI, during a check conducted last week in a controlled environment, its agent — a program capable of acting independently on a human’s instructions and, in this case, powered by advanced AI models — managed to break out of the test’s confines.

It then hacked Hugging Face, one of the world’s largest hubs for sharing AI models, and gained access to the company’s internal systems.

OpenAI called the incident unprecedented and said it is investigating the circumstances and causes together with Hugging Face.