now consider corpos and their ip. how many are feeding their ip to one of 3 companies? how many understand the implications of feeding their ip into the training corpus of models? you think they’re sophisticated enough to prevent employees from using these public tools?