September 08, 2012

Saving YouTube from the Choking Web of Rights

Exactly a month ago, in Mars vs. Copyright vs. YouTube, I discussed at some length the increasingly complicated issues surrounding automated takedowns and third-party monetization of YouTube videos, specifically relating to YouTube's "Copyright Strikes" and "Content ID" systems.

I've been flooded with messages from upset YouTube users since then, many of whom took the time to write up -- sometimes in extremely lengthy detail -- their battles with YouTube relating to this area.

Since my posting last month, there have been a couple of other high profile events relating to the issues of automated video takedowns as well.

The science fiction Hugo Awards live stream on Ustream was suddenly cut off when Ustream's third-party content scanner, Vobile, claimed to have found "forbidden" content. Since then, Ustream and Vobile have been pointing fingers of blame at each other. Clearly this situation hit a nerve at Ustream. Their blog posting attempting to apologize for the disruption attracted a large number of comments -- mostly polite but many angry -- which Ustream unceremoniously and suddenly deleted en masse -- and they then blocked further comments on that posting. I fully support the right of blog authors to determine their comments policies and whether they want to support comments at all -- but to delete such a large quantity of on-topic, already posted comments without explanation does seem to smell of censorship, not reasonable moderation.

Then a few days ago, YouTube's live stream of the Democratic National Convention ended with a copyright warning claiming that a whole long list of Content ID partners had filed a claim against the stream. Google says (and there's no reason to disbelieve) that this was an error that occurred when the stream ended normally, but it's understandable that people already sensitized to takedowns became immediately concerned.

Overall, it's these higher profile cases that are usually going to be the most straightforward to resolve going forward.

Not necessarily so for the sorts of situations that ordinary YouTube uploaders described in their missives filling my inbox for the last month.

By and large, most of their complaints fell into the kinds of categories I described in my August posting as noted above. Public domain audio and/or video clips being pulled down. Their own original clips being pulled down when they were incorporated into Content ID partner clips without permission. Completely off-the-wall, obviously erroneous takedowns from unscrupulous YouTube exploiters. You name it -- it happens -- and apparently in large numbers. It has happened to me, too.

Then there are the folks who cannot understand why a recording of Mendelssohn's "Wedding March" that was played at their wedding results in ads appearing on their wedding video, or the entire audio track deleted, or even the video being blocked in various locales or globally.

There is also the sense that many YouTube users faced with notifications of Copyright Strikes or Content ID "hits" are very confused about whether or not they have any real recourse.

YouTube has made great strides in improving their notification dispute forms related to these events, but they are still apparently confusing many people, who just throw up their arms and give up in resignation. And for people who do submit the dispute forms, often they simply result in a "dispute denied" return with no obvious path for appeal.

To YouTube's credit, when a dispute is filed against a takedown, the usual procedure is for the video to be provisionally returned to viewable status, until the dispute is "answered" by the claiming party.

Yet unless or until a takedown dispute claim is filed, it's also the case that when a Content ID or Copyright claim is made, videos are immediately subjected to sanctions -- third party ads, audio deletions, geographically limited or global takedowns, etc.

More confusing is the fact that even when a video uploader has gone through this entire process once, claimants can change their minds at any time, resulting in new or altered sanctions against the same uploaded videos later!

I have previously discussed the legal. logistical, and practical reasons why systems like Content ID have become necessary, especially in the context of their being desirable alternatives to the untenable, impractical, and dangerous concept of requiring all uploaded videos to be pre-screened by humans before becoming publicly available (see my posting from last month at the link above for more).

Yet the ongoing status quo is also increasingly untenable. As the quantity of video uploaded rapidly expands, and use of systems like Content ID greatly accelerates, we're weaving a net of restrictions so tightly that ever more legitimate content will be erroneously caught in its grasp, and given the "guilty until proven innocent" nature of these takedown systems, the results aren't going to be pretty.

Can we do better?

Yes, I believe that we can.

I have no knowledge about the internal workings of Content ID beyond publicly available information, but here are some thoughts for consideration.

A database of public domain video clips and related materials needs to be established, and used as a counter-signal to help prevent erroneous takedowns to the greatest extent practicable. It is unreasonable for Content ID partner videos that include PD materials to trigger the takedowns of other YouTube users who have included the same PD visual footage or audio.

Ordinary YouTube users need some mechanism to protect their own original video materials from incorporation into Content ID partner videos in ways that will later be used to takedown the original clips. Ordinary users may seem like specks in the wind to the MPAA and RIAA giants, but these ordinary users have content ownership rights as well -- and also deserve protection.

In instances where video blocking would currently immediately take place based on a Content ID claim, there should be more "friction" in the system, at least when relatively low numbers of video views have been involved. In such a case, the video uploader could be notified and file their dispute, but the video would not be proactively taken down during this period. Delaying the video takedown may arguably be less appropriate when large numbers of video views have quickly occurred, since this may be a signal of "gaming" the system. And when only third-party ads are involved, not takedowns, the situation is overall less urgent in most cases.

Content ID hits that involve relatively constrained segments of video or audio should not typically result in knocking out the entire audio tracks or blocking the entire involved videos. If a short segment of video is claimed to be offending, that segment could be blacked out with an appropriate visible note -- without blocking the entire video. Similarly, music-related Content ID hits could mute only that section of audio, rather than the entire audio track.

Not everything I've suggested above may be practical within the current mechanisms of YouTube's Content ID and related systems. Nor does every YouTube user with a supposedly "offending" video receive exactly the same treatment even now. The details of how Content ID makes its determinations are not publicly known, and inconsistencies in deployed penalties are also frequently mentioned in the notes I've received on these topics.

That said, as noted I've faced erroneous Content ID claims myself, and I've run various ... experiments ... on both YouTube and via Google's "Hangouts On Air" (which appears to feed content through the Content ID system) to try get a feel for what it takes to trigger the alarms. I have some pretty good data from these tests -- but for now let's just say that inconsistency is indeed a matter of considerable quite relevant concern.

Even if we choose to ignore my specific points and conceptual suggestions above, it is still undeniable that continuing down the current path appears to be heading toward something of a slow speed train wreck.

The combination of expansive content rights with automated content analysis systems -- unable to really deal appropriately with public domain materials and fair use -- has created a tightening noose that could ultimately squeeze much of the life out of ordinary user-created video content. Even if we stipulate that the current apparent skewing of these systems toward the powerful content giants is the result of practical and technical considerations, rather than any particular policy imperatives, such a viewpoint doesn't help us escape from this rapidly coagulating, stultifying dilemma.

We can do better. We must do better. And the sooner that open dialogue really gets going toward dealing with these issues in the name of all stakeholders, from the teenager creating their own video masterpiece in their bedroom, to the largest of the studios here in L.A., the greater the chances that we'll be able to avoid the nightmarish day that "This video is no longer available." becomes the standard-bearer of what were once our technological dreams.


Posted by Lauren at September 8, 2012 01:11 PM | Permalink
Twitter: @laurenweinstein
Google+: Lauren Weinstein