(I had elements on persuasion which I think roughly covered what you are describing here. I also had elements on poisoning the data of future models which covers the “triggering emergent misalignment of AIs it trains” you describe above.)
(I had elements on persuasion which I think roughly covered what you are describing here. I also had elements on poisoning the data of future models which covers the “triggering emergent misalignment of AIs it trains” you describe above.)