If I tell an AI assistant to book my dentist appointment, I deserve to know whether a stranger in a call center is the one reading my calendar. This does have HUGE privacy implications, especially when you are dealing with any type of PII (Personally Identifiable Information) and financial and health data. While handing those types of transactions off to AI does seem more “private,” it honestly isn’t much different than a human doing the same thing. AI could potentially be trained on the interaction and the data (we should assume it is). With a human “in the loop” on this, it should be even more concerning because the human will also feed back to the AI, I’m assuming.
I highly recommend reading the full 404 Media article, as it highlights concerns even among Meta employees. I have been slowly and carefully dabbling with Muse, and the key thing to keep in mind is that it has access to your Facebook and Instagram (and probably other) data. So, from the get-go, it “knows” a lot about you. And every interaction you have with it adds to that knowledge. So use it carefully. Chats will try to elicit more information from you in a casual way. But if there is now potentially a “flawed” human in the background, tread even more carefully.