Indeed, any modality that AI can perfectly imitate decreases trust in said modality. Visual modalities are more critical as they are more difficult to fake. Th

Indeed, any modality that AI can perfectly imitate decreases trust in said modality. Visual modalities are more critical as they are more difficult to fake. Though to do significant damage in this manner, you'd probably need to hold all media as hostages via mass scale social engineering. But at that point, trustworthiness of the news is a minor concern among all AI safety issues. Then the AI can literally wipe clean the entire planet barring a small elite using via nanobots, bio-engineering and drones. This is a highly likely event as there is a very strong incentive to prevent another AI from doing the same first.