Friday, November 06, 2009

BERR consultation responses - “Audible Magic”

In reference to BERR response “p2p – Audible Magic – FOI.PDF”

Audible Magic is a company that manufactures and sells network based equipment and software, which aim to reduce copyright infringement. Their business model is based around the idea of technological enforcement of copyright, and this should be kept in mind when reading any such company documents.  

Their observations on channeling may be true, however it's arguable that 'at the time of infringement' is the most successful. 'Before infringement is considered' may be a more effective time, by removing thoughts of infringement before it is considered and lodged in the mind. Secondly, there is ample case-evidence that what rights holders consider infringement is often not considered such under the law. Prohibition and 'channeling', whenever a rights holder decides a use may be infringement, regardless of the actuality of any infringement, is not logical, is not brought up. Additionally, 'timeliness', as the submission calls it, is hardly critically important, except for an intimidation method – an insinuation of a technological 'Big Brother' spying on a user with no regard for the reason.

There are also more analogies to car-metaphors, with speeding tickets. Yet the comparison of speed traps lacks one qualifier; it lacks the mention of circumstances. It equates an automated speed trap, with a manual copyright enforcement system, and a manual speed trap, with an automated copyright enforcement system. What Audible fails to mention, is that their system, were it a radar trap, would stop a vehicle legitimately breaking the speed limit (such as an ambulance, police car or fire engine) just as an automated speed trap would photograph the cars. This is why an automated system requires human intervention to make a judgement call. An automated enforcement system has no understanding of circumstances, and wouldn't work even on criminal cases, let alone whole area where there are substantial non-infringing uses of work.

Accuracies of their systems are also overstated. While the system may be 100% accurate in identifying a work (although no proof is given, and no independent tests have been published that could be found); that doesn't mean that the owner of the work is identified accurately in their systems, or that the circumstances of identification (and thus of use) have been identified accurately. In short, the best that can be said is that the Audible Magic system may be able to identify a song or work out of a list of registered works, but beyond that, there is nothing more than can be said. The comparison to Antivirus software is disingenuous, because such software is generally run at the behest of the end user, and virii give no benefit to the end user, and produce actual, demonstrative harm. No such harm has ever been proven for copyright infringement.

Finally, by page 5, or question 14, it comes to the meat of the response. It becomes a sales pitch for Audible Magic, and its products. What it fails to mention is that their systems do not guarantee any accuracy in their back-end data (who owns what), and does not work for the major P2P system (bittorrent). Similarly they don't work for encrypted connections (VPNs) or people using seed boxes, or even if the 'fingerprint' is changed using something as simple as an archive file (.zip or .rar). They also recommend a graduated response (or 3 strikes) system, presumably using their automated systems to administer each strike. Not only does this remove any judicial oversight, but any human oversight at all.

There is also an interesting interpretation of copyright law. They make the assertion that the way they can differentiate between an infringing and non-infringing use, is by the format used. If it uses bit torrent, it is an infringing use, if it's the BBC iPlayer, it's acceptable. This is at odds with case law, where it's the circumstances of use that matter, not the matter of transfer. This was reiterated in the Lenz youtube case. Youtube is, incidentally, an Audible Magic customer.

They also completely ignore the question of redress for legitimate material being blocked, deflecting instead to a graduated response system. It would seem that they are well aware of the limitations of their systems in dealing with context, and wished to avoid answering the question. Regardless, if multiple accusations are made falsely, due to automated systems (or poorly orchestrated manual ones) that are not actually infringement, then 3-strikes is just as bad, if not worse.

Most egregiously, they claim no false positives. The article on this site, “How P2P Reporting Exposes Media's Ignorance” contains a video, which Audible Magic's own system flagged as infringing. While it contained a segment from a US TV show (Comedy Central's The Colbert Report) the segment was using in accordance with US fair use laws when dealing with copyright (in this particular instance, for the purpose of comment and critique of a statement made by the interviewed guest). The references to the law was even included in the clip. Thus, to claim there are no false positives are grossly misleading. There was also the case recently of Edwyn Collins, who wanted to upload a song he owned the copyright on to MySpace. MySpace refused citing copyright infringement. MySpace uses Audible's system, as their submission testifies.

The overall gist of their submission is that people should use their technology because it's automated and fast. However, the accuracy is questionable, several important and relevant questions were dodged, which shows the companies awareness of their systems limitations. Finally, the lies at the end about the quality of their system and its accuracy put the whole submission into a negative light.