OpenAI says in their terms that you can't "use output from the Services to develop models that compete with OpenAI" [0], and it seems that people are interpreting it as training a model that directly competes with them, which we aren't doing. There are many companies out there built on using GPT-4 outputs to do task-specific fine-tuning, so it doesn't seem like it's a problem unless we were trying to make competing foundational model from GPT outputs.

(I'm not a lawyer, so this needs to be taken with a large grain of salt)

0: https://openai.com/policies/terms-of-use

It seems like everyone is doing it. Does anyone care? Should anyone care?