--- title: "Formalities" slug: formalities url: https://listedarticles.com/articles/formalities canonical_url: https://pluralistic.net/2026/10/31/burn-berne/ content_type: opinion language: en published_at: 2026-10-31T15:15:59.000Z updated_at: 2026-10-31T20:11:57.096Z author: "Cory Doctorow" authored_by: human publisher: "Pluralistic" publisher_url: https://pluralistic.net/ topics: ["ai-policy", "policy", "opinion"] license: all-rights-reserved word_count: 2074 reading_minutes: 9 citation: "Cory Doctorow, Pluralistic. \"Formalities.\" 31 Oct 2026. https://pluralistic.net/2026/10/31/burn-berne/ (all-rights-reserved)" # The full text follows. The web page shows an extract and sends readers # to the source above; quote the citation and link the canonical URL. --- # Formalities > Cory Doctorow argues the US Copyright Office is right that AI-generated works lack copyright, and that requiring human creative effort restores copyright to something you have to work to get rather than an automatic grant. # Formalities It's not often that I find myself on the same side as the US Copyright Office. In fact, I think this is the first time in 25 years that I've found myself cheering on the office's official actions as good for art *and* good for free expression, taking the side of creative workers over our bosses. The source of our miraculous accord: the US Copyright Office's position on AI. The CO has taken the (entirely correct) position that copyright law exists to incentivize human acts of creative expression, which means that only works of human authorship are eligible for copyright. This isn't a controversial position within legal circles. But for media executives, this is shocking; absurd, even. This gap between the industrial and the legal understanding of copyright comes from a fundamental difference over the evocatively named legal doctrine of "sweat of the brow." "Sweat of the brow" is fundamental to a correct legal understanding of copyright. Copyright rewards and protects *creativity* but it does *not* protect *hard work*. You can labor for ten years over a phone book listing the numbers of every person in a town and (as a collection of uncopyrightable facts) it will receive *no* protection *at all*. Zero. Zilch. Anyone can copy and sell that phone book: Copyright law doesn't care if you slaved over that collection of facts, if you missed sleep and your kids' little league games to realize your grand vision. Copyright law doesn't care if you plowed every cent into it, remortgaging your home and cashing in your (already resentful) kids' college funds. But: if you dash off a haiku in ten seconds while sitting on the toilet, shove it into your pocket and forget about it, and then your (still resentful) kids discover that crumpled haiku after your (highly suspicious) death, your kids will be able to claim a full *70 years'* worth of copyright over that forgotten triviality (the term of copyright for works of human authorship is life plus 70 years). This is the uncontroversial and universally accepted *legal* understanding of copyright: all the copyright in the world for creative acts, no copyright at all for mere acts of labor. The "sweat of the brow" has nothing to do with a work's eligibility for copyright. But no media boss believes this. For media bosses, copyright is a pseudo-property† right, designed and operated to protect their bottom line. The point of copyright isn't to incentivize *creativity*, but rather *investment*. † Or not-so-pseudo There's obviously a relationship between creative acts and acts of investment, but it's possible to create without investment (see: your haiku) and it's possible to invest without creativity (see: your phone book). And once again, the creativity that copyright protects is *human* creativity. This is a lesson that David Slater learned the hard way. Slater is a professional wildlife photographer and amateur monkey-trainer who combined his interests by teaching Celebes crested macaque monkeys to take selfies. But when Slater published a monkey's selfie others copied and distributed the image without his permission, on the (entirely legally correct) basis that monkeys are not humans, and that means that works created by monkeys don't get a copyright. This was also the Copyright Office's position, and the position of the Ninth Circuit Court of Appeals: This is the basis for the US Copyright Office's repeated decision to refuse to extend copyright to AI-generated works. An algorithm is even *less* human than a monkey, so if you can't get copyright over a monkey selfie, you *definitely* can't copyright your chatbot's slop. The US Copyright Office has successfully defended this position in multiple courts, triumphing over plaintiffs who were so persistent that they even got the Supreme Court to weigh in on the issue (the Supremes decided the case didn't warrant their attention): AI slop isn't copyrightable, but there can be copyrightable *elements* in an AI-generated image. The most obvious element is the prompt. So long as your prompt contains some quantum of creativity (e.g. "Make me a beautiful flower whose loveliness is reminiscent of a newborn baby", but not "make me a beautiful flower"), that prompt *is* copyrightable. That means you can use copyright to challenge others' rights to present the same prompt to a chatbot. But while the prompt can be copyrighted, the image it gives rise to *isn't*. So you can spend a month iteratively refining your prompt, consuming all your town's water along with 200,000 gallons of diesel and that prompt will be yours for your life plus 70 years, but the image it generates is in the public domain, and anyone can copy it, sell it or give it away. There are other ways that copyright can find its way into AI-generated material. You can make creative edits to it: if you output a slop novel and rewrite various sentences (providing those rewrites involve some creativity and not just mechanical corrections or word substitutions), then the edited sentences become copyrighted works, even though all the rest of the text remains uncopyrightable. You can also get a "thin" copyright through selection. If you get your chatbot to output *many* images based on your prompt and choose your favorite, your selection can confer a *tiny* bit of copyright over that creative act of selection, but it will be harder to defend than even the other (attenuated and highly specific) forms of copyright. Now, perhaps you've spotted an administrative problem lurking in this turn of events. How can someone else know what part of a work is AI-generated (and thus in the public domain) and which parts are human-edited/contributed (and thus copyrighted)? Just as importantly, how can *you* prove which parts of your work are not AI-generated and thus eligible for copyright protection. This gets especially gnarly when you're dealing with complex media. In an animation, an AI might do the rigging and movement for a human-created sprite. The sprite is copyrightable, but its movements aren't. That would mean that I could remap your sprite's movements onto a different sprite without infringing your copyright. Or say the sprite and its movements are human created (and thus copyrighted), but the backgrounds are procedurally generated? I could take those backgrounds and put new animations overtop of them. The simplest, most reliable way to ensure the works you invest in are fully and entirely eligible for copyright is to just use humans for the whole project. Declare a "no-AI policy," and hire human artists to produce the works you hope to sell. Monitor their activity to ensure they're not using AI. Keep good records. Now, as much as I like the idea that we've arrived at a copyright law that insists that a company can only get exclusive rights to its products by paying human artists to make them from end-to-end, I don't think that's likely to be the way it shakes out, but that doesn't mean that this is a wash for artists. The reality is that copyright claims are *always* "fact-intensive inquiries" that are expensive to litigate and difficult to adjudicate. It's completely *normal* for copyright lawsuits to turn on documentary evidence gleaned from fragmentary records. This is simply intrinsic to copyright, because copyright is a doctrine that requires judges to assess whether things are "creative" and "original" as though these were freestanding categories that everyone could agree upon. Of course, the rules of copyright can make this harder or easier. One way in which the rules make deciding copyright cases *much, much harder* is copyright's broad prohibition on "formalities" ("formalities" is how copyright wonks refer to "red tape" or "paperwork"). For nearly all of American history, copyright ran on a key formality: *deposit, registration and notice*. To secure a copyright for a new work, an artist (or their employer) had to send a copy of that work to the Library of Congress (deposit), along with paperwork detailing the claim of authorship (who made this, what parts of it did they make) (registration) and they had to display a copyright *notice* that told you who held the copyright to every work that was registered (notice). Both the filing and the notice required careful attention and precision, and screwing either one up could result in your losing your copyright. It's an open secret among copyright scholars that Walt Disney screwed this up with the first Mickey cartoons, meaning they've been in the public domain since their release: This system of copyright formalities meant that you could always tell whether a work was copyrighted, you could always find out who claimed the copyright over it, and which parts of the work were claimed under copyright and under what basis. What's more, formalities created the Library of Congress, since the LoC acted as America's "copyright deposit" library, holding deposit copies of every work that had ever been issued a US copyright. The US system of formalities was descended from the English system, in which the Stationers' Guild used a copyright registry to establish who had the rights to which books. But there were competing copyright systems, notably the "continental" system that was codified through the Berne Convention, a copyright treaty drafted by Victor Hugo (yes, *that* Victor Hugo). Among other things, the Berne Convention abolished formalities. Much of the world adopted Berne, but the US held out until 1976. With the 1976 Copyright Act, Congress brought the US into compliance with Berne, abolishing formalities in the USA. As a compromise, the 1976 Act requires US authors to register their copyrights in order to obtain "statutory damages" (up to $150,000 per work). That's a pretty good incentive, but even with that incentive, the vast majority – far in excess of 99.999999999999999999999999999999999999999% – of all copyrighted works are not registered with the Copyright Office. They are born copyrighted from the moment they are "fixed in a tangible medium" until 70 years after their authors die. But without registration, you can't know when that copyright clock started ticking because you have no way to verify who the author of the work is. This not only makes it *incredibly* difficult to establish the copyright status of nearly every copyrighted work, but it also means that in nearly every case, you can't readily determine who you should approach if you want to pay them for a copyright license. Abolishing formalities has created a universe of orphan works: everything from napkin doodles to kindergartners' finger paintings to message board posts. Within that universe is a kind of gaseous cloud that, if condensed, would come out to an infinitesimal planetoid, representing the entirety of commercial works whose authors give a damn about their copyright status. It's a bad, bad system. This system doesn't do away with the need to keep careful records: like I said, copyright disputes usually require the close examination of documentary evidence. What formality abolition accomplished was the destruction of the regular system of record-keeping: rather than having a standard, universally agreed-upon way to determine the status and resolve disputes about a copyrighted work, now we have the scattershot, incoherent system that makes copyright disputes *far* more likely (because no one knows who to approach for copyright licenses) *and* far more difficult to resolve (because there's no registry of copyrighted works). Which brings me back to the question of AI and copyright. If a firm uses *some* AI in its work, it's going to need to document that *very* carefully. Firms that don't keep very close track of the elements of their products that come from AI will have a far harder time defending their commercial interests. Basically, we're about to recreate formalities, but managed by the private sector. To the extent that companies use AI in their work, they are going to have to track every pixel and every word and every note that is output by an AI system, and which parts were added or edited by humans after creation. That means that companies that want to protect their commercial interests aren't just going to get more protection based on how much work they get humans to do; they're also going to have to treat copyright's government-enforceable monopoly over expressive speech with the gravitas it deserves: not something you get automatically, but something you have to do real administrative work for. ---