My suggestion is that humans are still noticeably better at this, not least because humans are able to backtrack and re-adjust their interpretations, which LLMs

My suggestion is that humans are still noticeably better at this, not least because humans are able to backtrack and re-adjust their interpretations, which LLMs cannot in the same way. An LLM's "thought process" and "output" are essentially the same thing. Now maybe I'm wrong about this, but even if so I think it's still a risky change. We have understood human error for pretty much all of recorded history. We have a very poor understanding of computer error, to the point that there are some legal systems where computer software is assumed to be correct by default. Courts and laws around the world are designed to cope with humans making mistakes and lying.