Like me, you find the Open Source AI Definition weak on the training data information side? You'd be right and there's a reason for it... it's probably hiding quite some open washing for the larger models. This is a good explanation of the motives and consequences.