Generative AI Forces Copyright and Antitrust to Collide

Copyright and antitrust have long operated in parallel, rarely intersecting in meaningful ways.

Copyright law governs incentives for creativity and authorship. Antitrust law focuses on market power and exclusionary conduct. For decades, the two doctrines largely stayed in their own lanes.

Generative artificial intelligence is changing that separation.

That was the focus of a recent TechREG interview with Daryl Lim, H. Laddie Montague Jr. Chair in Law at Penn State Dickinson Law and Associate Dean for Research and Strategic Partnerships. Speaking with Competition Policy International, a PYMNTS company, Lim said generative AI introduces the need for copyrighted works at industrial scale.

“When you train frontier models, you need to ingest vast repositories of works that may include copyrighted works,” Lim said. “Only a handful of firms can do this, and those firms often control the compute, the data, the cloud infrastructure and distribution simultaneously.”

That concentration is what causes copyright and antitrust to collide.

The Paradox of Bigness in AI Markets

The dynamic is a paradox of bigness, Lim said. The same scale that makes AI systems more powerful, reliable and safe also raises concerns about dominance and entrenchment.

Historically, copyright disputes rarely implicated antitrust because copyrighted works were fragmented and substitutable. Control was dispersed across many rights holders, making durable market power unlikely. Generative AI breaks that logic. Training large models requires aggregation. That aggregation naturally favors a small number of vertically integrated platforms.

The risk is that dissatisfaction with monetization or licensing outcomes begins to substitute for proof of competitive harm, Lim said. If courts allow that shift, antitrust risks becoming a proxy for copyright enforcement rather than a tool for addressing exclusionary conduct.

“The harder but more coherent approach is to let copyright’s internal safeguards do their work, while reserving antitrust for cases that really show demonstrable exclusion rather than platform innovation,” Lim said.

Why Fair Use Sits at the Center of the Debate

Much of the public debate around generative AI assumes that training models on copyrighted material is clearly unlawful. Lim said that assumption oversimplifies the legal questions.

Fair use has long functioned as copyright’s internal competition safeguard. Courts have relied on it during past technological transitions involving photocopying, search engines, reverse engineering and software interoperability. In each case, the doctrine was used to distinguish learning and transformation from substitution.

“What matters is not that AI is new but whether the doctrine is structurally suited to distinguish learning from substitution,” Lim said. “And we have an entire body of jurisprudence for that.”

That does not mean the legality of AI training is settled. Courts in the United States, the United Kingdom and Europe are actively weighing questions around non-expressive use, secondary liability and the mechanics of training. But Lim cautioned against layering antitrust enforcement on top of unresolved copyright questions.

If conduct that copyright law ultimately permits is later deemed exclusionary under antitrust, firms face conflicting commands. Markets do not become more competitive. They become frozen.

Conduct, Not Size, Should Trigger Antitrust

Scale alone should not trigger antitrust intervention, Lim said. Scale is intrinsic to AI development and often improves performance, safety and reliability. Antitrust has long grappled with that tradeoff, most notably in earlier technology cases involving software integration.

The guiding principle remains conduct, not size.

Antitrust scrutiny is warranted when there is demonstrable foreclosure. Lim pointed to examples such as exclusive cloud or compute arrangements that deny rivals access to essential training resources, data partnerships that restrict access on unreasonable terms, or coercive placement contracts that reduce market contestability.

Those are familiar antitrust concerns grounded in exclusionary harm. By contrast, disputes over whether training constitutes infringement or fair use belong squarely within copyright law.

The Risks of Politicized Enforcement

Beyond doctrinal boundaries, Lim raised broader concerns about the politicization of antitrust across administrations and jurisdictions. When competition enforcement becomes a tool of ideology rather than evidence, predictability suffers.

“For markets to be investable, you need legitimacy, neutrality and predictability,” Lim said. “It doesn’t matter who is in power. When enforcement becomes unpredictable, it can erode confidence and innovation.”

He warned that antitrust is increasingly being asked to address issues better handled by other bodies of law, including industrial policy, labor concerns and expressive outcomes.

Competition law is well-suited to evaluating market definition and foreclosure, Lim said. It is not designed to resolve broader social or political objectives.

Why Regulatory Clarity Matters Most

Lim’s paper outlines a five-part framework for navigating AI-related regulatory conflict, including regulatory clarity, compliance by design, institutional reform, policy realignment and empirical research. If policymakers could act on only one, Lim said regulatory clarity should come first.

Clear separation of institutional roles allows each body of law to do its own work. Copyright governs authorship, infringement and fair use. Antitrust addresses exclusionary conduct and foreclosure. When those mandates blur, enforcement becomes politicized, and innovation slows.

Predictability matters as much as enforcement itself, Lim said.

“When those domains are clear about their respective mandates and neither is pressed into the service to resolve the other’s core questions, innovators can invest with more predictable boundaries,” he said.