It's not a 0-1 thing, there are degrees of attraction towards "evil". The guard in a concentration camp has more tendency towards evil than the guard in, say, a factory or residential building.
As Buchheit[1] himself said: "It’s also a bit of a jab at a lot of the other companies, especially our competitors, who at the time, in our opinion, were kind of exploiting the users to some extent." (emphasis mine)
Realizing that your competitors, i.e. firms you acknowledge as being in the same industry, are "kind of exploiting the users" seems to me not far from realizing that your very industry has a tendency towards that.
I think "Don't be evil" is an honest expression of the early confidence and idealism of the founders and core team of Google. However, in the last few years, as with America at large, something seems to have changed at Google. Personally, I tend to steer more and more away from it.
If they truly wanted to avoid evil and, perhaps more important, the potential for future evil, they would anonymize all observations to make, say, search per se better (as opposed to expressly attributing data so as to "personalize" it). That, incidentally, would also make Google immune to NSA bullying. Much like Apple, they could just say: We don't even store this kind of data, sorry guys, look somewhere else. At the rate with which they collect data points and with which hardware, algos, and science advance, it will be possible in a few years to do fine-grained psychological profiling on a massive scale, with all the possibilities of evil I can imagine. It's a fine line between prediction and control in social systems. And we all know that what is technically possible eventually will be tried by someone. It's not hard to think of scenarios.
_________________
[1] http://blogoscoped.com/archive/2007-07-16-n55.html