I considered using it for a Spring project a couple of years back, and opted against it because it was a tremendously heavy and complex framework and the flexibility that it provides just wasn't necessary IMHO. It was (in my estimation) less effort to roll our own authentication/authorization. Don't misinterpret this as meaning that it was a trivial effort; effective security never is.
From a risk standpoint, I didn't understand it deep down after spending some time with the documentation, and decided that the complexity represented a significant risk of misconfiguration. It may be "better" than what we built, but if we didn't understand how to use and configure it properly, then it wasn't going to live up to its potential. A custom-implemented (and possibly "inferior") security module that I understand inside-out is less concerning.
Disclaimer: Spring Security was still called Acegi at the time, and the current technology may well have changed along with the name.