It has subsequently been improved by Gary Robinson and Tim Peters, among others.
[2] The most notable difference between a conventional Bayesian filter and the filter used by SpamBayes is that there are three classifications rather than two: spam, non-spam (called ham in SpamBayes), and unsure.
If the scores are both high or both low, the message will be classified as unsure.
This approach leads to a low number of false positives and false negatives, but it may result in a number of unsures which need a human decision.
Some work has gone into applying SpamBayes to filter internet content via a proxy web server.