I know nothing about coding, but I have been a very active Wikipedia editor for 20 years.
I understand that ChatGPT uses Wikipedia articles as training data, but not the edit histories of those articles, which show the various iterations that each article has gone through to reach its current state. It is a very rare thing to have that data for any substantial piece of writing, and I can think of few data sets that would provide a more instructive example of how to write by adding good information, removing bad information, and restructuring existing content for better presentation.