They can't even reliably answer questions about their own training!

no, I also find the concept of "ai lying" to be incorrect as they lack intent

I could not tell what parts the human author thought were phantasies they felt they had escaped versus which they really were wallowing in. Who honeypots the honeypots?

No I don't get it at all. But I think a lot of people are prompting LLMs then publishing their outputs these days. They need to jump on the hype train before it leaves the station. See any recently created "AI ethics" group/institute/startup/whatever. It's basically just this meme on repeat:

https://knowyourmeme.com/photos/3229293-say-i-am-alive