In theory it's a crowd-sourcing effort: if enough people report a site as problematic for one reason or another, this state gets triggered, and scary messages are displayed. The reason doesn't have to be actually true. If enough people says it's true, it's good enough. Just like real life.