For what it's worth, they seem use blocks as large as 128 KiB, if I understand correctly. Do you know how likely an overlap of such huge blocks is? Complete novels and pictures fit easily into such a huge block. Considering this I don't see their point that this is supposed to prevent copyright infringement at all. Do you think 20 seconds of an song are not copyrighted? Artists have been sued for sampling less than a second of someone else's work. Even if you could assume that this is "Fair Use", keep in mind that "Fair Use" exists only in US law and it might not be applicable for this case anyway.