… and yet it’s extremely misleading for ChatGPT to guessing urls that don’t exist when asked for references! I had the same happen when asking it a series of questions about laws and policies: it cited a policy I hadn’t heard of, so I asked for a link reference and it provided a very suspect link that went nowhere. That’s a really big caveat emptor! Unfortunately it’s already out in the wild a misinforming the majority who hasn’t read the fine print as closely. The minimal responsible thing for OpenAI to do would be to put disclaimers inline in chat whenever chatgpt generates a non-existent “reference”