I’m curious about this too! One thought is that if I’m using a secure AI that’s integrated in the tools that I already use and trust (like Dovetail), then it would be less essential to inform participants about than if we were dumping transcripts into ChatGPT