Some things can’t be blamed on human error. Computers make mistakes, too.
Roughly 30% of requests to takedown copyrights material are of questionable validity, according to a new study by the University of California, Berkeley, and Columbia University, and one in 25 of these requests targeted material that wasn’t related to the works they allegedly ripped off.
To crunch the numbers, researchers looked at a random sample from more than 100 million notices sent by media companies and the agencies they tap to keep the movies, music and porn they produce from getting uploaded, shared or streamed illegally. Because these companies depend on automated, “bot”-based systems to police the Internet, it’s leading to problems. The algorithms they deploy aren’t too hot at discerning between content theft and parody or determining what differentiates piracy from fair use. There are laws in place, after all, that allow news organizations and others to send up or comment on copyrighted materials.
“You have computers making decisions,” said Jennifer Urban, a professor of law at Berkeley and one of the study’s co-authors. “It’s hard to put the fair use doctrine in a computer chip.”
These software errors add up — nearly a third of online takedown requests had flaws and would have benefited from human review, the study found. The music industry was responsible for the greatest percentage of takedown notices (41%), followed by adult entertainment (28.1%), movies and television (17%), software companies (7.5%), and game businesses (5.4%).
Part of the problem may be that the system of checks and balances put in place by lawmakers may be fraying. These takedown requests trace their origins back more than a decade to a 1998 law, called the Digital Millennium Copyright Act (DCMA), that was passed in order to resolve copyright infringement disputes during the Wild West days of the Internet. But for online companies like Google, the number of these requests has ballooned from thousands a year to tens of millions a week amid an increased reliance on automated systems of enforcement.
The challenges posed by piracy are real. A 2013 Columbia University report found that nearly half of U.S. citizens pirate movies and television, and also discovered that with younger consumers, age 18 to 29, that number climbs to nearly 70%. And having a movie leak online can be financially disastrous. A 2014 report by Carnegie Mellon found that revenues for films that were pirated before they debuted dropped nearly 20%.
The problem is that the system being used to protect copyrighted material may be getting abused. Although there are flaws with the bot systems used by entertainment conglomerates, the Berkeley and Columbia study also cites problems with the less technologically advanced notices sent by individuals and small businesses. The report looked at takedown notices related to Google Image Search, which tended to be submitted not by robots, but human beings. Seven out of 10 of these requests were of questionable validity, and they disproportionately targeted blogs and personal websites, potentially infringing on freedom of expression.
“There are a lot of issues about a lack of sophistication regarding copyright law and when it is appropriate to ask for things to be taken down. Some people are misusing the takedown process because it’s super-cheap,” Urban notes.