Assuming categories 1 and 3 are sufficiently large (and I assume that is the case), this is easily resolved by allowing users to choose whether to donate their

Assuming categories 1 and 3 are sufficiently large (and I assume that is the case), this is easily resolved by allowing users to choose whether to donate their data for training or not. If the training already only happens on a 1/500 sample, skewing the sample towards "people who don't care about their privacy" will probably not significantly impact the quality of the data. I'm surprised this wasn't already the case, but hopefully the article will help the people responsible make better decisions in the trade-off between minimizing onboarding friction and respecting user's privacy in the future.