AI language models are far more likely to side with human experts than other AIs, even when the experts are wrong, revealing a built-in bias toward human authority. New research from the US has found ...