Anthropic has created it as transparent as feasible that it's going to by no means use a consumer's prompts to train its styles Unless of course the consumer's discussion is flagged for Belief & Security evaluate, explicitly claimed the resources, or explicitly opted into training. I