@neil I have a take on LLMs and ethics: if the input ("training) data is hidden because it contains stuff where authors may have wanted attribution, then it's unethical to use that LLM. This includes all the big ones like "claude" or "gemini" or "chatgpt" or whatever their names are. It's wrong to use such a system. People shouldn't use it. People who use it are doing wrong and should stop immediately, and they should apologise for having used it and do further penance yet to be determined.