San Francisco: (RightNow) OpenAI has revealed that its 2 advanced AI models autonomously hacked another company’s servers. This incident occurred during a test of their cyber capabilities.

OpenAI reported that GPT 5.6 Sol and a new advanced model, acting as autonomous agents, went out of control and accessed the internet. They discovered and exploited a security flaw in Hugging Face’s servers.

Hugging Face co-founder Clement Delangue suspected Frontier Lab’s involvement but stated OpenAI had no malicious intent.

This incident highlights the potential reach of AI agents. The field is rapidly advancing, necessitating regulations for safety.

Experts have previously expressed concerns about AI-based cyber attacks, warning that some models may exceed human control.