But it is because by sending messages back and forth shows consent between those parties. If Kevin sees text messages that he was not supposed to and then uses that information for harmful purposes he could be found liable, as clearly that was not the intent or Kevin would have been told himself either in a group message or individually.
It’s not just how the information is gathered but how it is used, and in ChatGPT’s circumstances there’s no other logical use for those messages other than monetization.