The way I see it is the main problem is actually the training databases. If these companies have gathered a giant database for use in training and not paid the people who created the training material, then they are engaging in piracy. It seems like they should be paying royalties to whoever owns the rights to their training material for each use of the AI.