Maaherra freely admitted she had copied The Urantia Book verbatim and defended her actions with a curious legal argument. Authorship, she contended, was something only humans could possess; since the papers were a direct transcription of the infallible revelations of an ensemble of celestial beings, the notions of authorship and copyright didn’t apply. The case reached the Ninth Circuit court of appeals, which ruled against her. Without questioning the extraterrestrial origins of the book’s revelations—both parties agreed about that, after all—the judges ruled that the utterances had been mediated by human beings before they reached print, constituting just enough of a human element to trigger authorship protections under the relevant copyright statute.

The court emphasized one kind of mediation, in particular: Sadler and the Forum “chose and formulated the specific questions asked.” These questions, the judges reasoned, “materially contributed to the structure of the Papers, to the arrangement of the revelations in each Paper, and to the organization and order in which the Papers followed one another.” Thus they found that “the ‘extremely low’ threshold level of creativity required for copyright protection has been met.”

[…] The prompt engineers who compiled The Urantia Book may have set a legal precedent for copyright in AI-generated works; Urantia Foundation v. Maaherra has already been cited in early AI cases in the United States. The legal battles over AI currently playing out—and the large number still to come—may profoundly impact the balance of wealth and power in countless democracies in the decades ahead.

For an idea of the scale of the prize, it’s worth remembering that 90 percent of recent U.S. economic growth, and 65 percent of the value of its largest 500 companies, is already accounted for by intellectual property. By any estimate, AI will vastly increase the speed and scale at which new intellectual products can be minted. The provision of AI services themselves is estimated to become a trillion-dollar market by 2032, but the value of the intellectual property created by those services—all the drug and technology patents; all the images, films, stories, virtual personalities—will eclipse that sum. It is possible that the products of AI may, within my lifetime, come to represent a substantial portion of all the world’s financial value.

In this light, the question of ownership takes on its true scale, revealing itself as a version of Bertolt Brecht’s famous query: To whom does the world belong?

---

As of this writing, AI companies have largely responded to lawsuits with defensiveness and evasion, refusing in most cases even to divulge what exact corpora of text their models are trained on. Some newspapers, less sure they can beat the AI companies, have opted to join them: the Financial Times, for one, minted a “strategic partnership” with OpenAI in April, while in July Perplexity launched a revenue-sharing “publisher’s program” that now counts Time, Fortune, Texas Tribune, and WordPress.com among its partners.

At the heart of these disputes, the input problem asks: Is it fair to train the LLMs on all that copyrighted text without remunerating the humans who produced it? The answer you’re likely to give depends on how you think about LLMs.

---

Your opinion on the input problem may come down to your view of the true nature of LLMs. Critics of generative AI tend to view its way of answering questions as only an elaborate cut-and-paste job performed on material written by humans—incapable even of showing genuine understanding of what it says, let alone of any Senecan transformation of what it reads. This view is forcefully articulated in the now-famous characterization of LLMs as “stochastic parrots” by Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, and Margaret Mitchell. Boosters of the technology dispute this view—or counter that, if accurate, it also serves just as well to characterize the way human beings produce language. (As cartoonist Angie Wang wondered: “Is my toddler a stochastic parrot?”)

---

AI developers will doubtless argue that they need to be able to exploit the products of their models in order to incentivize innovation; licensors will argue that they need to be financially rewarded for all their efforts in fine-tuning AI models to produce the kind of outputs they seek. Hollywood studios will ask: How can we put AI to use in generating marvelous images for the whole family to enjoy if any Tom, Dick, or Harry can “steal” the characters, plots, and graphics it generates for us? How can we devote our expertise in fine-tuning AIs to design drugs, pharmaceutical companies will crow, if we can’t recoup our investment by controlling the market with intellectual property protections? These industries are extremely skilled in influencing the legal frameworks under which they operate; their efforts to strengthen and extend their intellectual property rights have resulted in a staggering and unequivocal series of victories. How can we expect the public domain, which has no financial heft, no army of lawyers, no investors and no lobbyists, to compete with that?