For your Pilates example, many people here would say that proper prompt-engineering is needed. After all, the AI doesn't know which pilates class you meant. Perhaps, the AI would go to his home town's pilates center and book a spot there, even though he is half a world away. Perhaps, the ai would book a spot in the lecture with Professor Pilates.

So, yes. That person could have some culpability. After all, if you don't expect your middle school child to perform the task as instructed, maybe you should reword it before tapping Send.

You can go to ANY AI and put in "so what are the various meanings of book a spot at the pilates class ? In other words, what might an ai do with that instruction, which I argue is poor prompt engineering." You might be surprised at how poorly your AI thinks that is worded.