> it's already known that their model is very likely trained on input gathered from scientists which use OpenAI as a tool for their research

Is this actually known or just broadly suspected? I could never quite figure it out...