Whenever an Internet user searched the Web, attached a file to an e-mail or examined a menu of files using file-sharing software on a peer-to-peer network, the software would compare the hash values of those files against the file registry. It wouldn't be "reading" the content of the files -- it couldn't tell a love note from a recipe -- but it would determine whether a file is digitally identical to one on the child-porn list.

Imagine browsing for Blu-ray dumps on one of these monitored p2p-networks, and for each file in the search result, GFR would download it from the other user, read it, and discard, just to compute a hash value. Afaik., making a hash of a file involves reading the data. It seems that the author of the article (or CopyRouter itself) differentiates between opening and reading.