Current LLMs don't pass the turing test.

Remember, the interrogator is allowed to be hostile, so they would obviously employ all known prompt injections and typical LLM 'gotchas' to figure out who the AI is.