[dmd]

>> It looks like this isn't something I can help you with. Would you like to be connected to a human who can help?

> Sure!

>> Ok, I'm connecting you to a human now.

[5 minutes later]

> Hello?

>> Hi! What can I help you with?

> Are you a human?

>> No, I'm an AI agent programmed to help you with anything you need. What can I do for you?

> You said you were going to connect me to a human.

>> That isn't something I can do. What can I help you with?

Turns out "connecting to a human" is something it knows about in its training data so it'll hallucinate doing so.

[dylan604]

At least your example IDs itself as an AI agent. The ones I've come across hide it but it becomes obvious with responses like "I don't have access to that information" or something a human would never say. I had a dealership give me one of those, so I hung up on it and called a different dealership where I was connected to an actual human. Guess which one got my business...