I think OP is more concerned with we are going to ensure that new data added to the "trusted corpus" is going to be free of LLM taint.