I guess it can be assumed that a good number of people read the PKGBUILDs, so at some point malware would be found. A peer-reviewing system would give people a false sense of security, since the AUR is a user repository, where breakage should be expected (compared to the official repos).