The internet’s decline, referred to as the “enshittification” by experienced journalist and privacy advocate Cory Doctorow, traces its roots back to a time well before TikTok entered the scene. Those belonging to the elder millennial generation can reminisce about the days of Napster, which was soon followed by the dark era when Napster and similar P2P sharing networks were embroiled in legal battles, leading to their demise. What remained were a handful of streaming platforms approved by record labels, offering sanitized music catalogs, such as Pandora and Spotify. Hooray for corporate copyright litigation.
In his recently released book, “The Internet Con: How to Seize the Means of Computation,” Doctorow delves into the current social media landscape, dissecting and illustrating the numerous shortcomings and shortsighted business decisions made by the dominant tech companies that promised us a bright future but instead delivered more problems, including the proliferation of extremist ideologies. Doctorow contends that we not only have an obligation but also a responsibility to deconstruct these systems, emphasizing the potential for greater interoperability.
In this excerpt from “Hitting the Books,” Doctorow explores the aftermath of the legal actions against P2P sharing services and dissects the roles played by the Digital Millennium Copyright Act’s “notice-and-takedown” reporting system and YouTube’s “ContentID” scheme in shaping the landscape of modern streaming platforms.
Excerpted from by The Internet Con: How to Seize the Means of Computation by Cory Doctorow. Published by Verso. Copyright © 2023 by Cory Doctorow. All rights reserved.
Seize the Means of Computation
The implications of notice-and-takedown procedures, on their own, do not directly impact major entertainment corporations. However, in 2007, the entertainment industry devised a more potent variant of notice-and-takedown that not only directly affects Big Content but also exacerbates the consequences for all of us.
This innovative system is known as “notice-and-stay-down,” which succeeded the notice-and-takedown approach. It monitors all content uploaded or generated by users and cross-references it to identify any similarities with previously flagged copyrighted works. The entertainment industry had long pursued this legal objective, and it became a part of EU law in 2019. In 2007, it was first introduced as an optional modification to YouTube, known as “Content ID.”
Some context: In 2007, Viacom, a subsidiary of CBS, filed a billion-dollar copyright lawsuit against YouTube, alleging that the platform had actively encouraged its users to infringe on its copyrighted content by uploading it to the site. Google, which had acquired YouTube in 2006, defended itself by invoking the principles established by the Betamax case and the notice-and-takedown procedure. Google argued that it had fulfilled its legal obligations and that the Betamax case had determined that tech companies could not be held liable for copyright infringement induced by their services (recall that Sony had marketed the VCR as a tool for breaching copyright laws by recording and sharing Hollywood movies, and the Supreme Court had ruled that such marketing didn’t matter).
However, given the looming threat of Grokster, Google had reasons to doubt the viability of this defense. There was a real risk that Viacom could litigate YouTube out of existence, particularly since internally obtained communications from Viacom, filled with strong language, showed that Viacom executives had been debating who would gain control of YouTube within their organization when Google was compelled to sell YouTube to another company.
Google narrowly secured a victory but was determined not to face another situation like the Viacom lawsuit. In response, Google developed Content ID, an “audio fingerprinting” tool presented as a means for copyright holders to either prevent or monetize the use of their copyrighted content by third parties. Initially, YouTube allowed prominent copyright holders to upload their content catalogs to a blocklist, which Content ID would then scan to compare with user uploads, checking for any matching audio.
When Content ID identified a user attempting to upload copyrighted material without the permission of the copyright holder, it consulted a database to determine the copyright holder’s preference. Some copyright holders chose to block any uploads containing matching audio, while others opted to claim the ad revenue generated by the video.
However, there were numerous issues with this system. Notably, Content ID could not discern whether a third party’s use of someone else’s copyright fell under the category of “fair use.” Fair use includes a range of uses permitted even if the copyright holder objects, such as using excerpts for critical or transformative purposes. Determining fair use often requires a nuanced consideration of the specific circumstances, making it a “fact-intensive” doctrine. In most cases, the question “Is this fair use?” would lead to the answer, “It depends; let’s consult a judge.”
Computers are unable to distinguish fair use from copyright infringement, and it’s unlikely they ever will. Consequently, filters tend to block various legitimate forms of creative work and expressive speech, particularly those involving samples or quotations.
Filters encounter difficulties not only with creative borrowing, remixing, and transformation but also with similar creative works. For example, a six-note phrase from Katy Perry’s 2013 song “Dark Horse” closely resembles a six-note phrase in “Joyful Noise,” a 2008 track by a lesser-known Christian rapper named Flame. Perry and Flame engaged in legal battles, with Flame accusing Perry of copyright infringement. Perry ultimately prevailed, which was a positive outcome for her.
Nevertheless, YouTube’s filters struggle to distinguish Perry’s six-note phrase from Flame’s, and filters often perceive copyright infringements that don’t exist, even when the similarities are not as pronounced as in the example mentioned. This is an intentional feature of the system.
To comprehend this, we must view filters as a security measure—a strategy employed by one group of individuals (platforms and copyright holder organizations) aiming to prevent another group (uploaders) from engaging in an activity they wish to perform (uploading infringing material).
Creating a filter to block precise matches is a straightforward task. Copyright holders could upload digitally pristine, losslessly encoded master copies of their entire catalog. Any user attempting to upload a track identical in digital or acoustic terms to the master copy would be automatically blocked.
However, users can easily circumvent such a filter by making slight, imperceptible audio alterations that take the uploaded file just below the threshold of human perception. Alternatively, they can trim a fraction of a second from the beginning or end of the track, omit a single musical bar from the bridge, or apply a myriad of other modifications that listeners are unlikely to notice or raise objections about.
Filters do not rely solely on exact matches. Instead, they employ a “fuzzy” matching approach. They don’t just block content that copyright holders have specifically identified; they also block content that bears a resemblance to what copyright holders have claimed. The degree of fuzziness in matching can be adjusted, allowing the system to become more or less strict in determining a match.
Copyright holder groups prefer matches to be as broad as possible because there may be someone who is content with a very fuzzy, truncated version of a song, and they wish to prevent that individual from accessing the song for free. Looser matching criteria result in more false positives. This poses a particular challenge for classical musicians, as their performances of works by composers like Bach, Beethoven, and Mozart naturally sound quite similar to recordings claimed by Sony Music through Content ID. Consequently, earning a livelihood from online classical performances has become nearly impossible, as videos are either blocked or the ad revenue they generate is diverted to Sony. Even teaching classical music has become a hazardous endeavor, as meticulously crafted free online lessons are either blocked by Content ID or, if the label permits them to remain online, the ad revenue they generate is funneled to a major corporation, effectively robbing music teachers of their creative earnings.
The notice-and-takedown law did not deliver the internet that copyright holders had envisioned. In reality, their actions, along with revealing memos from the Viacom case, indicate that combatting copyright infringement serves as a pretext for establishing an internet where entertainment companies dictate who can create new technologies and how they will operate.