I do get that impression from some of the crash descriptions. The cars are extremely cautious and will often brake in situations where normal drivers wouldn't. There was one, for example, where a Google car was rear-ended while very slowly passing an accident site with emergency vehicles on the emergency lane (they were not on the road itself). Many drivers wouldn't even slow down, or only slow down a little bit, while the Google car did slow to a crawl. Other drivers simply don't expect this, they are passing the site at normal speed, probably getting a peek of the accident, and all of a sudden there's this "idiot" (in their eyes) Google car in the way. There was also one where a Google car was crossing an intersection, with right of way, a car from the other street rolled through a stop sign, the Google car braked because it judged that there would be a "near" collision, and the other car hit the rear end of the Google car. Quite possibly, the other car's driver had judged that he could pass behind the Google car and did not expect it to suddenly brake. The near collision turned into a real one.
There were several examples like that. And yes, I know what a lot of people will say, driving slowly is safer, you should always slow down when there's an emergency vehicly anywhere in sight, you should never roll through a stop sign, etc... Fact is, though, that most human drivers have certain expectations about other drivers, and the Google car often behaves differently. Sudden braking "just in case" is not always a good idea. You may say afterwards "yes, the human reacted too late, he wasn't paying attention, not keeping enough distance, going too fast, etc...". But if it does indeed turn out that Google cars are in more accidents (even though they are never at fault), that may mean that some adjustments in its logic are in order. Simply to accomodate the mistakes and bad behaviour of humans.