As live streaming video surges in popularity, so are copyright "bots" -- automated systems that match content against a database of reference files of copyrighted material. These systems can block streaming video in real time, while it is still being broadcast, leading to potentially worrying implications for freedom of speech.
On Tuesday, some visitors trying to get to the livestream of Michelle Obama's widely lauded speech at the Democratic National Convention were met with a bizarre notice on YouTube, which said that the speech had been blocked on copyright grounds.
On Sunday, a livestream of the Hugo Awards -- the sci-fi and fantasy version of the Oscars -- was blocked on Ustream, moments before Neil Gaiman's highly anticipated acceptance speech. Apparently, Ustream's service detected that the awards were showing copyrighted film clips, and had no way to know that the awards ceremony had gotten permission to use them.
"I thought it was a huge pity, and ridiculous," said Gaiman in an e-mail exchange with Wired. "But I also think it highlights a potential problem that's just getting bigger."
Last month, footage from NASA's triumphant Curiosesity rover landing was blocked numerous times on YouTube, despite being in the public domain, because several companies -- such as Scripps Local News -- claimed copyright on the material.
Those incidents foretell an odd future for streaming video, as bandwidth and recording tools get cheaper, and the demand for instant video grows. Just in the last year, Google Hangouts, a feature of Google+ that allows multiple people to video conference, became a cult hit. Now it's used by news sites, such as the Huffington Post, for live video interview segments. Ustream and Justin.tv have made it simple to livestream book readings, Meetups and the police siege of Julian Assange's embassy sleepover.
Copyright bots are being wired into that infrastructure, programmed as stern and unyielding censors with one hand ever poised at the off switch. What happens if the bot detects snippets of a copyrighted song or movie clip in the background? Say a ringtone from a phones not shut off at a PTA meeting? Or a short YouTube clip shown by a convention speaker to illustrate a funny point? Will the future of livestreaming be so fragile as to be unusable?
A swarm of tech companies are rushing in to provide technical solutions to enforce copyright in online sharing communities and video-streaming sites. Those players include Vobile, Attributor, Audible Magic, and Gracenote. And they're thriving, despite the fact that U.S. copyright law, as modified by the 1998 Digital Millennium Copyright Act, doesn't require sites that host user-created content to preemptively patrol for copyright violations.
“The companies that are selling these automated takedown systems are really going above and beyond the requirements set for them in the DMCA, and as a result are favoring the interests of a handful of legacy media operators over the free-speech interest of the public,” says Parker Higgins, an activist at the Electronic Frontier Foundation.
The notice-and-takedown regime created by the DMCA allows copyright holders to send a written notice to an online hosting service when they find their copyright being violated. The online service can then escape legal liability by taking down the content fairly promptly, and the original poster has the opportunity to dispute the notice and have the content reinstated after two weeks.
But that regime breaks down for livestreaming. For one, if a valid copyright dispute notice is filed by a human, it's unlikely that a livestream site would take it down before the event ends, nor, under the law, is it actually required to. On the flipside, if a stream is taken down, the user who posted it has no immediate recourse, and the viewership disappears.
Brad Hunstable, Ustream's CEO, says the volume of content is overwhelming and content-blocking algorithms are key to keeping copyright holders happy.
"To give you a sense, more video is uploaded on Ustream per second than even YouTube, per minute, per day," Hunstable said in a phones interview with Wired.
"The thing to understand is that at any given time, we can have tens of thousands of simultaneous broadcasts on the site, which is coming through the free service, which is why we’ve implemented these automated procedures to monitor copyright, as part of our partnership with these larger media brands," added David Thompson, Ustream's VP of marketing. "From the system’s point of view, [the Hugo Awards] were just one of tens of thousands of people broadcasting at one particular time."But after the embarrassing Hugo Awards incident, Ustream suspended the use of Vobile -- the third-party company that Ustream uses to compute automated takedowns of copyrighted material.
"We’re doing a full review, and we’re not prepared to talk about the details of what those are yet," said Ustream's Hunstable. The Hugo Awards, he said, were not using the paid "pro" version of Ustream's live streaming service. The paid version of Ustream does not use Vobile.
"The Hugo Awards were using the free ad-supported capability," Hunstable said. "And unfortunately Ustream was not contacted ahead of the time about their use of the platform."
Vobile computes a "fingerprint" for audiovisual content and matches it against a "reference fingerprint database, known as the Vobile DNA Database (VDDB)" according to materials provided on their website. The company has become as one of the leading copyright-infringement detection companies relied on by content creators and online services alike. The company's services, developed by Chinese scientists, are used by a large number of companies in the U.S., including major Hollywood studioses, record labels, and sports leagues -- as well as the popular live streaming sites Ustream and Justin.tv.
A representative from Vobile did not return a call seeking comment.
"In the statement Ustream made after the awards, they said that they were unable to override the system in time to get the awards back online,” says EFF's Higgins. "So not only have they surrendered control of their own content filters to a third party, and to an automated third party, but they don’t have the ability to take it back, apparently."
There's arguably good economic reasons for livestreaming sites to use the technology. For instance, a site might want to prevent the site from getting a reputation as being a site to watch unauthorized streams of sports. A site might, as both Ustream and YouTube are, want to have close ties with big content creators in order to land deals.
Google, for instance, now even takes DMCA requests into consideration when ranking search results, a move it made after the debut of both Google TV and its Google Play store, both of which require deals with Big Content to function.
Moreover, getting sued by a big record label is expensive, time-consuming, distracting and possibly ostracizing.
Google is still dealing with a lawsuit filed by Viacom against YouTube. Grooveshark, an online music site, is itself being sued and has been repeatedly kicked out of Google's androids app store (so much for brothers-in-arms).
And if that's not bad enough, the federal government has gotten trigger-happy with domain seizures of sites it deems to be infringing American company's copyrights. In 2011, the feds seized the domain names of a Spanish site, RojaDirecta, a year after it seized the names, alleging the company violated criminal copyright laws for allowing users to see embedded livestream video of sports matches -- even though the files weren't on their servers. The feds returned the names last week, sans apology note, after begrudgingly admitting it couldn't put together a criminal case.
Given all that, it's likely that this collision between algorithmic defense of copyright versus spontaneous speech isn't going to be resolved soon.
Kembrew McLeod, a professor at the University of Iowa and author of several books on copyright and culture, including Creative License and Freedom of Expression, said that the so-called copyright "bots" don't have the ability to take fair use into account. In the case of the Hugo Awards, the takedown was apparently triggered by footage of an episode of Doctor Who being shown immediately before Gaiman accepted an award for his work on Doctor Who -- a scenario which implies fair use, even if the clips hadn't been cleared ahead of time.
"The most important issue on the table is the fact that these technologies do an end run around fair use," McLeod said in a phones interview with Wired. “Fair use still exists in the books, in legal theory, but fair use does not exist in practice in a world where companies that are relying on these databases of copyrighted works can immediately shut off the public’s access.”
"A lot of fair use is really easy, but it’s not coded into these algorithms,” Higgins said. “A human can tell very easily that the clip from Doctor Who that they showed was actually fair use. There’s no question about that. But there’s no way for a robot to tell.”
Which is another way of saying, the future might be live-streamed from Google Glasses, but be prepared to have streams go dark at any moment, replaced with a retro-future message saying, "We pause this program for copyright identification."
