Even if you had pure provenance of each response, it still will not stop large tech companies from abusing the public's desire of "not knowing". Consider searc

Even if you had pure provenance of each response, it still will not stop large tech companies from abusing the public's desire of "not knowing". Consider search engines as an old example. It took multiple decades to even provide a concept of a "why am I seeing this?" affordance. That only became a recent topic after majority of people started caring about the legitimacy of the news as their values also changed. Even then, it isn't even helpful nor provides actual context based on your search history or advertisement profile. Why? Because keeping people in the dark makes great advertisement targets. When people become aware of why they were targeted, it's like holding up a mirror revealing our major flaws. We don't want to identify with certain things although an algorithm may slap the label on us. We are ashamed of such labels and want to immediately distance ourselves from them. I believe the exact same will be injected into these models to the point where you will be suggested consumeristic practices when you're just trying to accomplish something else entirely. So in a sense, you'd be manipulated and hijacked of your original intent. Search engines and social media do a great job at this already today. Now that there's a strategic shift across the industry, it is especially unlikely that any provenance will be provided for trained models. How exactly will you know that the model you're using isn't biased with certain leanings? > Regulators must consider this an urgent danger. How will regulators even know nor care if they don't see it already happening today?