To be fair to your concerns, how do we deal with the problem of propaganda in AI? How do you correct for institutional bias in de-biasing results? How do you gu

To be fair to your concerns, how do we deal with the problem of propaganda in AI? How do you correct for institutional bias in de-biasing results? How do you guard against censorship that can itself push minorities out of being able to use those models? Well, you correct for that by having diverse models trained by diverse people that aren't localized to a singular company or insulated within a single industry. Which is... also kind of a big thing that Mozilla is arguing about here. User agency over models and training, privacy and local models, having a diverse set of companies and organizations building models rather than a single monopoly that controls access and training data, transparency and inspectability of AI that allows us to examine why it's producing certain output -- those are the actual ways to protect against de-biasing efforts turning into propaganda, and Mozilla's points above are decent ways to get closer to that goal. Those are the important steps you should be focusing on rather than policing companies for acknowledging and trying to reduce harm against minorities.