That seems a bit disingenuous (not that I think you're wrong with respect to Google slurping up data, but that's already been discussed at length and is mostly orthogonal to differential privacy). A perfectly valid use case for differential privacy is to take the enormous pile of data Google has access to and construct models and views into it that don't leak anything "too personal" in some rigorous sense to the outside world. IME they tend to use more crude techniques like not showing any Google Trends data if there aren't enough data points, but whether Google uses this kind of technique at all is a much more general question than whether they use it to protect the public from Google itself.