Earlier quoted context omitted.
Would they need to delete derived ML data as well?
My understanding is that as long as there is no PII is in the ML model (there’s not) then any existing models do not need to be deleted.
It has been proven countless times that it's possible to extract learning data from models. I can't see how you can prove the opposite, except, maybe, with federated learning (but even then, you need to good "ratio" of noise)