Afaik all the large providers flipped the default to contractually NOT train on your data. So no, training data context size is not a factor.