The goal of an ethics team will be to provide cover for problems created by decisions or suggestions of an llm. They will need to be able to say "We trained it to consider ethical considerations by " <including something like all the reddit and usenet postings ever>. We tested by <some other inadequate strategy that sounds good to lawyers>.

They will use "industry standard practices", they will have tested it somehow, and will be sure to be on the board of some group that is building toothless "ai ethical tests" for llms.

I agree, which is a different job from what I expect most AI ethicists imagined it would be.