The AI copyright watershed comes into view

Last year — I went back and counted — fully a dozen of these predictions concerned generative AI, a topic I am convinced most Americans hadn’t even heard of at that time. A year later, that’s only still widely true among those frequenting caves. But comparatively little of the attention around news and AI has focused on its impact on the future economic viability of our industry, and 2024 seems likely to be the year in which this changes.

Once upon a time, less than 20 years ago, the platforms gobbled up the vast bulk of advertising revenues, mostly by operating at previously undreamed-of scale, thus being able to deliver to marketers hyper-targeted audiences at very low prices. Search was a big part of this (including search for news), but publishers couldn’t bear to be excluded, even when Google offered them robots.txt as an opt-out. Like the man in the joke at the end of Annie Hall who couldn’t give up the imaginary chicken, they needed the eggs of referral traffic.

This time, however, having seen that movie (Publishing Armageddon, not Annie Hall), big publishers are refusing wholesale, uncompensated consent to the ingestion of their content by large language models. Whether they can make this stick — and force the platforms to pay in perpetuity — hinges largely on a question of copyright law, whether the products resulting from AI are protected as what the law terms “fair use.”

If they are, the platforms will likely significantly disintermediate news itself. You could ask OpenAI to tell you what’s new in the Israel-Hamas war, or the presidential campaign, or just what’s news generally without reference to the underlying sources. Ultimately, whether the law permits this — I happen to believe strongly that it doesn’t — is a question that is going to need to be decided by the Supreme Court. By the end of 2024, particular cases raising this question will be well on their way to the Court, and the issue should be sharply posed, in much the way the future of affirmative action was a year ago, or the future of abortion two years ago.

Luckily for the press, which probably isn’t all that popular at the Court these days, the implications go well beyond the news. All creative industries are at risk from a loss on this key copyright question. That’s because we are not long from a time when generative AI could, for instance, craft something that would look very much like a 34th Marvel Cinematic Universe movie or a new James Patterson novel. (Sherlock Holmes, by the way, as a creature of the 19th century, is in the public domain — look for new adventures, eerily evocative of Conan Doyle, before long.)

If, in 2025 or so, the Court rules in favor of underlying copyright holders on this question, major publishers will craft their own deals with AI providers in exchange for a small piece of the action (some are already doing so provisionally), and smaller publishers will likely have a choice of accepting a standard agreement and a small fee, or opting out. In that case, AI will still have enormous implications for publishing costs, work processes and capabilities, but the threat to the news business probably won’t be existential.

If the cases go the other way, however, an industry that is already marking its 19th consecutive year of business decline may find that its own news is capable of getting much worse.