[Vingdoloras]

CC-BY-SA is a perpetual license, so the author can't just decide that they don't want StackOverflow to have access to the content anymore.

[int_19h]

The point is that OpenAI is not sharing the result of that training with us, nor even the training set. In other words, they are ignoring both "attribution" and "share-alike" parts of the license, claiming that it's not applicable because what they are doing is fair use, and so copyright is not a concern.