Scalable systems need to use heuristics to catch threats. Needing concrete evidence in every case means that an enormously higher amount of malicious resources will not be flagged.
There is a policy argument as to the right balance of concerns here. But there is a clear trade-off to make.
