The Democrats just won an extraordinary wave election, taking strong majorities in the House and the Senate and winning most states’ gubernatorial races, too. At a high level, winning Dems ran on what some are calling an “affordability and accountability” agenda—rooting out corruption and lowering prices for ordinary people. As the folks who rode this wave into office prepare to act on their agenda, they should ignore bogus copyright monopolist rhetoric about “theft” and embrace fair use (and other limitations to copyright control) as a tool to promote fair competition and affordability.
As this Re:Create infographic shows, political candidates use fair use almost every day in an election season. Whether you’re quoting your opponent’s gaffe or a newspaper’s endorsement, fair use lets campaigns spread the word about issues that matter to voters. Even if you don’t know what fair use is, if your campaign just won an election, you probably used fair use along the way. A high-profile example comes from Scott Colom’s historic win in Mississippi, which was powered in part by a viral ad that included a fair use video clip of his opponent telling a news anchor that Mississippians upset about expensive beef should “pick another protein.”
Fair use isn’t just for politics. It’s a flexible right for anyone to use copyrighted works as part of a transformative, new creation, from critical essays to news reports, from documentary films to search engines and (according to two federal court opinions so far) artificial intelligence models. Crucially, fair use doesn’t require permission from (or payment to) a copyright holder. This ensures that creators of new art and new technologies can pursue their passion without facing a veto from the big incumbent media companies who control most commercial copyrights.
Another crucial limitation to copyright is its scope: copyright does not apply to abstract ideas, generalizations, facts, or other fundamental building blocks of expression and culture (like grammar or style). This ensures that anyone is free to build new products, services, and creative works based on what they learn from existing copyrighted works, as long as they don’t publish the work itself (or a copyrighted derivative, like a translation or adaptation). Edgar Allen Poe might’ve written the first detective story, but later novelists are free to try their hand at the genre without asking permission from the Edgar Allen Poe estate. One newspaper may be the first to report the cause of a major wildfire, but follow-on reporters are free to include that new fact in their own reporting.
That brings us to affordability. Fair use and the free circulation of ideas and facts empowers creators, publishers, and technologists to bring more options to the marketplace, increasing competition and helping to control prices on information and entertainment products. As founding fathers like Thomas Jefferson and James Madison understood, copyright may be a useful incentive to creativity, but it is also a monopoly, and monopolies make goods scarce, expensive, and crummy. That’s why the copyright monopoly has to be limited, including by fair use. Copyright limitations ensure that the copyright monopoly stays in its lane, protecting a rightsholder’s opportunity to sell or license their individual work without letting copyright holders block others from competing for consumer dollars with new works and new technologies.
These principles are very much at work in the AI economy. As antitrust and IP expert Michael Carrier explained last year in Lawfare, if you’re worried about monopolies in AI, you should embrace copyright limitations. AI models are “trained” by analysis of billions or trillions of “tokens” of data, representing words and phrases in existing published artifacts. That training encodes public domain facts, ideas, and generalizations into a model’s architecture, making these powerful elements of the public domain more readily accessible and useful to the public. While copying is an inevitable step in the process of gathering and analyzing data, fair use permits that copying because it is “non-expressive” – no ordinary consumer ever sees or uses the work in a way copyright regulates. The market that copyright protects—the consumers interested in paying for a particular work—is untouched. What changes is that there are more options available to the consumer.
Carrier identifies several reasons why protecting fair use for AI training is good for competition and affordability. First, if AI training data had to be licensed (which experts say is impossible, at least at the scale required for cutting edge models), the cost of licensing would be a major barrier to entry for start ups and competing AI companies, for non-profits and researchers (including AI critics), and for creators and others using AI in their work. Nothing would do more to crown the handful of current AI leaders as permanent winners than limiting fair use and requiring copyright licenses for training.
Fair use also protects consumers from the other big oligopoly in this space – the big copyright holders like the New York Times, the Big Five book publishers, and the three major labels, who control the majority of commercial copyrights. Without fair use, these mega-corporations would use their dominance to choose winners in this new technology and siphon off the vast majority of the proceeds of any licensing regime. (Just ask recording artists how Spotify works for them.) They could also ensure that AI technology can’t be used by independent creators in ways that offer consumers competition against corporate media products. If Big Content controls AI, Carrier explains, “Copyright holders might ignore productive uses that matter to society but yield limited revenue—for example, models trained for scientific research purposes, to study AI itself, or to engage in biomedical research.” Without fair use, Disney, Paramount, and a handful of foreign-owned music and book publishers would be effectively put in charge of a general purpose technology that could lower prices or bring new services to the economy in ways we haven’t even imagined. Copyright was never meant to convey that much control.
Some news publishers are going even further, asking for control over facts themselves. They want to squash competitors who offer news summaries based on the facts in publicly available reports. Re-reporting news that first breaks elsewhere is a valid and useful practice employed by every news outlet, enabled by the bedrock principle that no one owns facts. But big news publishers like the New York Times would like to change that, suppressing the free circulation of facts in order to shore up their dominant position. As Re:Create member R Street has argued persuasively, this is not a recipe for a vibrant news landscape; it’s a path to increased consolidation and oligopoly, further erosion of local news in favor of a few national and international conglomerates.
Voters sent a message in this election: they want a fair deal, not a rigged system where a few big companies call the shots and pad out their profits with prices normal people struggle to afford. The copyright oligopolists have gotten a free pass from the media and in the public imagination by framing themselves as Davids against a Big Tech Goliath, but the reality is that they are Goliaths, too, and their copyright portfolios already give them extraordinary power in the marketplace. They wield that power at the expense of consumers and working creators. Fair use and the public domain ensure that content oligopolies can’t leverage their ownership of past culture into control over our creative and innovative future.