I spoke recently with someone who is building such a website/community, and I doubt they will mind me giving a brief overview of how they intend it to work.
Basically there are 2 issues that are difficult to overcome - assuming all the technological issues are minor:
Skepticism in the sense of 'Oh, I see this PI has a bad review from 2001, but who's to say that will happen to me. Maybe this one student is an odd-ball'. It also has parallels with the TornadoGuard problem.
And we see this a lot on academia.stackexchange too. We only ever hear 1 side of the story, and that story is rarely impartial. You know, occasionally people like to rant. People often omit the times they made mistakes. It's human nature. And so any website of this nature has to somehow overcome this issue of he-said-she-said, and theres simply no technological device we can implement to get around this. Particularly if you allow anonymous submissions.
The second, slander, is what I believe most people in the pubpeer.com debate are worried about. A deliberate attempt to hurt someone's reputation under the flag of mob-justice. Nitpicking and emotive words, but no real substance, etc. I actually only found out about pubpeer a week or so ago when a colleague at a party showed it to me, and I was shocked to see people I respect being picked apart on there for silly little things. Still, those anonymous criticisms do have some validity, and the concept of post-publication peer review is something I deeply agree with. I think only time will tell what impact sites like pubpeer will have on science. For a PI-review site, it could have disastrous consequences. Potentially legal consequences.
The website/community I referred to up top seemed to have spent a considerable amount of time fleshing out all the technological methods that could help here, and in the end i'm told they settled on the fairly low-tech idea for Skype-based interviews, potentially on a weekly release schedule. The idea is to mainly cater for people who hate their PhD and feel relief knowing they're not the only ones going through hell. So these Skype interviews generally are with post-docs or no-docs who had a bad experience, and they talk about what the end result was, what they'd do differently, if they still feel they were treated unfairly many years down the line (I watched 3 interviews and they were all totally different, so i'm finding it difficult to generalise here). But the big issue for the team putting all this together is finding PhD/post-docs/no-docs who are OK with talking about their experiences. They offered to interview me but I turned them down because, frankly, i'm not that brave. They offered to pixilate my face, but even so... as disappointed as I am with my treatment during my PhD, there's no one specific person that really let me down. It was a system of failures, I think, and one that is unlikely to happen again. So that particular PI-review project wouldn't really help me or people like me.