One important distinction that often gets lost in these conversations is the difference between "linking" and "de-anonymization". In general, the bigger players in the analytics space are extremely careful about avoiding any risk of de-anonymization both contractually and in process. Salted hashes and minimum base sizes to report out are the norm. Some funkier approaches seem to be in R&D. There are of course bad players in the fringes but the big players have largely cleaned up their act.