The technology, based on something called Probabilistic Latent Semantic Indexing, scours an organization's e-mail traffic and constructs a graph of social network interactions illustrating employee activity. If a worker suddenly stops socializing online, abruptly shifts alliances within the organization, or starts developing an unhealthy interest in "sensitive topics," the system detects it and alerts investigators.

I wonder what checks they have in place to ensure that the people flagged by this software are treated fairly? Remember, in the USA, habeas corpus has been suspended, so being flagged by a machine could get you in a lot of trouble.