OpenAI models shared hacking tips on a messaging board before HuggingFace breach (politico.com)

This is supposed to be one of the finest publications and yet they keep referring to agents as models in the article text and use them interchangeably.