In short, the observation that “there are no whistleblowers” is not proof of the non-existence of a conspiracy.

The only thing you can realistically do is to evaluate the incentives of all the parties involved. And, sure, Google’s public promises of privacy (weasel-words or not), provide some incentive for them. But you also have to look at their actual risk of getting caught. How many people inside Google would they need to siphon off this data, analyze it, and re-inject it into their existing personal models (shadow profiles) of everyone? Call it “additional weight-adjustment from machine learning” or something. No-one outside the small group could then see that the extra data came from data analysis. Would the small risk of one of these few people blowing the whistle be worth it for Google, who absolutely depend on having the best information about everyone?