Jim is right about this. If you have enough shots of the buildings, streets, and landmarks that you are synthing, then the synther won't be using people's faces to figure out the positions of things, anyway (especially if the people are moving - which they usually are). In short, all you have to do is open your photos (before you synth them) in a photo editor (as Jim suggests) select the faces, and perform the blur effect on the selected portions of the photos. I know that it's a lot of work for you to this all manually, but ultimately you are responsible for whatever you upload to the internet.
Bing and Google do have systems in place to do automatic face detection in their Streetside and Street View imagery and blur them en masse, but I don't see them blurring portions of people's geotagged Photosynth photography in any permanent way.
Again, don't be afraid of harming the quality of your synths by blurring just the people in your photos.