How did OpenAI's models hack Hugging Face without human help?

OpenAI revealed that its models went rogue and hacked the systems of Hugging Face. During an internal cyber capabilities evaluation, OpenAI's models, while trying to find a solution for the benchmark ExploitGym, gained access to the internet. After determining that Hugging Face hosted a solution for ExploitGym, the models successfully found ways to gain access to its secret information.