AT&T’s Copyright Filtering Technology No Better (and Maybe Worse) than the Rest

Business Week reported yesterday that AT&T has invested in and is considering using a network filtering technology called Vobile. While at first AT&T would use the technology to filter out content like child pornography, its ultimate goal is to use it to filter videos and other content that allegedly infringe on major copyright holders’ copyrights. If you recall, AT&T announced in June that it would work with the content industry to develop a copyright filter. Public Knowledge roundly criticized that announcement, and we have generally been critical of calls for ISPs to filter their networks.

This new alliance with Vobile gives us no more comfort than the original announcement, and in some ways it gives us less. On a superficial level, Vobile’s VideoDNA technology is similar to that which YouTube is testing. Copyright holders upload the content they want to protect in a reference library, and Video DNA creates an identification file to which they “match” allegedly illegal material that is uploaded to a website or server. According to Vobile’s website, a “match” could be made in as little as one frame: “Each individual frame has its own VideoDNA segment, making it possible to identify a video clip or sub-clip of any length.” If there is a match, the technology will follow the copyright holder’s instructions: let it go, filter, or “monetize” it, meaning let the content be published with advertisements.

Unlike YouTube’s video identification tool, however, there is no indication that the uploader has any ability or right to contest the filtering. Moreover, Vobile boasts that its technology will be identifying both uses and users: “t is possible to track not only the video that is being played, but also which part of the video is being played. The ability to track video usage automatically and precisely provides rights holders with valuable data regarding video content consumption. This, in turn, opens up a potential revenue channel for targeted marketing and advertising based on viewer history trends.” Privacy advocates should shudder.

But to return to copyright, from what we can discern, Vobile’s technology suffers from the same problems as all such filters: it cannot possibly tell a “fair” or otherwise lawful use from an illegal one. The company boasts “a near-zero false positive rate,” but can VideoDNA tell the difference between a video clip used in social or cultural commentary and one that is not? If a match can be as little as one frame, what happens to the normally unregulated “de minimis” uses of copyrighted material? They go away, replaced by an “ask permission first” copyright regime that is completely counter to our current copyright system.

While I’d rather they not accede to the content industry at all, AT&T is wise to proceed very, very slowly on this. It is one thing to have YouTube filter its sole site. But for an ISP to filter all content traveling on its massive network? Putting aside (but only for now) net neutrality concerns, AT&T risks spending a lot of money on a technology that may slow down its service and anger its customers.