|
Going further, the court cited Campbell for the proposition that the “potential market” includes both the market for the original work and markets for derivative uses. According to the court, evidence showed that “the market for licensing headnotes as text to train AI is rapidly developing” and that Thomson Reuters was already using its headnotes as training data for its own AI search products. The court then added an important point for emerging markets generally: “That Thomson Reuters did not license its headnotes to others does not disprove that a market exists to do so.”
The court put the matter directly:
ROSS’s copying harmed Thomson Reuters in this market. By copying the headnotes for use as training data without Thomson Reuters's authorization, ROSS usurped Thomson Reuters’s opportunity to enter that derivative market and license its headnotes for that purpose. The market harm caused by ROSS's use weighs against fair use.
The court’s factor four analysis in ROSS is especially noteworthy because it recognized, in the context of AI technologies, that copyright law protects developing as well as mature markets. Campbell teaches that potential derivative markets include those that copyright owners “would in general develop or license others to develop.” In its ROSS decision, the court also relied on Castle Rock Entertainment, Inc. v. Carol Publishing Group, Inc. (1998), for the proposition that a potential derivative market is not illusory merely because the copyright owner has shown little interest in exploiting it. Castle Rock itself put the point more broadly: “copyright law must respect that creative and economic choice.”
ROSS and Generative AI: Copyright Principles for Emergent Licensing Markets
The significance of the court’s reasoning in ROSS extends beyond the particular market at issue in the case, especially where unauthorized use itself can impede development of a lawful market. Courts should not authorize widespread uncompensated use of copyrighted works that weakens emerging licensing markets and then point to the resulting lack of licensing activity as favoring fair use. Fair use should not create the market failure later invoked as evidence that no copyright market exists.
Campbell’s recognition that potential markets include those copyright owners reasonably would develop or license others to develop is complemented by Warhol’s first-factor warning that an overly broad conception of transformative fair use could “swallow the copyright owner’s exclusive right to prepare derivative works.” Taken together, those decisions underscore that a use may differ from the purpose for which the original work was created and still implicate a valuable market that the copyright owner may choose to enter, license, or decline to exploit. Different purpose does not mean no market.
The facts in ROSS concerned AI training for a legal-research platform. But analogous licensing markets are foreseeable wherever generative AI systems use expressive works – including books, journalism, photographs, music, sound recordings, and audiovisual works – to produce expressive outputs in the same or related media. The broader takeaway is that factor four protects more than sales of existing works; it can also protect a copyright owner’s opportunity to participate in new AI-training markets by developing its own products, licensing works to others, or choosing not yet to exploit the market. The court’s reasoning thus provides appellate-level support for a central proposition: the economic value of AI-training rights should not be presumed to belong to unauthorized users who exploit the new technology first.
The court’s recognition in ROSS of an emerging AI-training licensing market also fits comfortably within the constitutional property-rights foundations of American copyright law. A central part of copyright’s value is the owner’s exclusive authority to decide whether, when, and on what terms others may use the work. In its ROSS decision, the court applied that principle in the AI context by recognizing that a rightsholder may itself enter a new AI market or license others to do so, and that an unauthorized user can impair or usurp that opportunity.
More broadly, where new uses of copyrighted expression create valuable licensing opportunities, those opportunities should properly belong to the copyright owners of the works. Voluntary AI-training markets should be allowed to develop rather than be distorted through expansive fair use rules. Copyright owners may use their works in their own AI products, license others to use them, or choose not to participate, while AI developers remain free to negotiate for the inputs they value. Such exchange allows prices, contractual terms, and efficient licensing mechanisms to emerge, while the prospect of compensation supports continued creation and dissemination for the public’s benefit. By contrast, declaring a developing category of AI uses free at the outset risks effectively extinguishing copyright owners’ opportunity to enter a new market.
* Seth L. Cooper is an Adjunct Senior Fellow of the Free State Foundation, a free market-oriented think tank in Potomac, MD. The views expressed in this Perspectives do not necessarily reflect the views of others on the staff of the Free State Foundation or those affiliated with it.
| 
