The Line Between Influence and Infringement
In the last few years, generative artificial intelligence has evolved from a novelty into a genuine dilemma for the music industry. Tools like Suno, Udio, and Anthropic’s Claude can now compose melodies, lyrics, and even full songs that sound remarkably similar to existing works. Because these models are trained on massive collections of preexisting material—much of which remains under copyright—they sometimes reproduce familiar fragments or stylistic patterns that echo the originals. That overlap raises a difficult question: at what point does inspiration turn into imitation? The boundary between lawful influence and copyright infringement has always been hazy, but generative AI has made it nearly invisible. As new lawsuits emerge, the issue now sits squarely at the intersection of art, law, and technology. How courts decide these cases—whether AI outputs count as transformative or derivative—will determine not only how music is made, but who receives credit, profit, and authorship in the age of algorithms.
Borrowing ideas in music is nothing new. Every musician builds on what came before—there are only twelve notes spread across eighty-eight keys– so repetition is inevitable. Jazz improvisation, blues chord progressions, and hip-hop sampling all rely on creative reinterpretation of earlier work. Copyright law was written to allow that kind of influence while still protecting originality, but AI has made that balance harder to sustain. Unlike a human artist who listens, learns, and reshapes ideas, an AI model processes information by deconstructing millions of songs, lyrics, and recordings into statistical patterns. What it produces isn’t conscious creativity but a mathematical reconstruction of existing art. That distinction matters because it changes what “influence” really means. What used to be a lineage of inspiration is turning into an algorithmic feedback loop—something copyright law was not designed to regulate.
Copyright law itself was built to protect human creativity by giving artists control over their work. The first American copyright statute, the Copyright Act of 1790, granted authors fourteen years of protection (renewable once) for works like books and maps. [1] Later laws, especially the Copyright Act of 1976, expanded coverage to “original works of authorship fixed in any tangible medium of expression.” [2] Once an idea takes a concrete form—like a recorded song or a written lyric—it is protected.
To determine when one work unlawfully borrows from another, courts apply what’s known as the substantial similarity test, first established in Nichols v. Universal Pictures Corp. (1930). [3] The test asks whether an ordinary listener would recognize that the defendant’s work copied protected elements of the original. For example, in Bright Tunes Music Corp. v. Harrisongs Music, Ltd. (1976), George Harrison’s “My Sweet Lord” was found to infringe The Chiffons’ “He’s So Fine” because the melodies were almost identical. Harrison said it was unintentional, but he was still ruled against. That case made it clear that intent doesn’t matter: If the similarities are strong enough, there’s infringement.
There are still exceptions. Under 17 U.S.C. §107, a work may qualify as fair use, a doctrine meant to balance creative borrowing with protection. Here, judges consider at multiple factors: purpose, nature, amount, and market impact—but the most influential is whether the new work is transformative, meaning it adds new meaning, message, or purpose. In Campbell v. Acuff-Rose Music, Inc. (1994), the Supreme Court ruled that 2 Live Crew’s parody of “Oh, Pretty Woman” was fair use because it changed the tone and message of the original. Since then, “transformation” has become the key question in copyright law disputes. For decades, courts have relied on that standard to separate copying from creativity. But all of it rests on one core assumption: that there’s a human author behind the work, making intentional creative choices.
That assumption collapses when it comes to artificial intelligence. Generative systems are trained on enormous datasets of music and text, with much of it being copyright material. The training process itself involves copying data to analyze it, and that reproduction lies at the heart of copyright protection. This issue came to a head in Thomson Reuters Enterprise Centre GmbH v. Ross Intelligence Inc. (2025), when a federal court found that an AI company had infringed by using a proprietary legal database to train its model without permission. [4] The court rejected the argument of being “transformative,” emphasizing that large-scale ingestion of copyrighted materials can still violate exclusive rights if it doesn’t serve a fundamentally new purpose. Although that case dealt with legal texts rather than music, its reasoning easily applies: training a model on copyrighted recordings to generate “original” songs may itself amount to copying.
That ruling exposed how copyright law is not updated for machine-made art. The system was built on human intent—someone choosing to copy or to create. An algorithm doesn’t make that choice; it just predicts patterns. Yet its predictions can recreate melodies, voices, or lyrics identical to the real thing. The music industry is now grappling with two new major issues: record labels suing AI developers for infringement, and artists using AI to push creative boundaries. The law can punish reproduction and reward transformation, but it struggles to define either when the “artist” is not human.
Even so, AI isn’t always a threat to originality. Many musicians use generative tools the way they use instruments like synthesizers or samplers. When an artist directs a model, alters its results, and integrates them into a new composition, the outcome likely qualifies as a transformative work under existing law. But when a system autonomously generates a song that closely mirrors a preexisting melody or lyric, it stops being an instrument. The real challenge for courts will be drawing that line clearly without stifling artistic innovation.
For that to happen, copyright law will need to evolve. One possible path is a compulsory licensing system, similar to what already governs sampling and radio airplay. AI developers could train on copyrighted material by paying standardized fees to rights holders, compensating creators while allowing innovation to continue. Another reform could be transparency requirements that force companies to disclose what works are absorbed in their training datasets, giving artists the ability to monitor or opt out. Both approaches would preserve copyright’s core purpose—protecting creative labor—while acknowledging that machine learning is–and will continue to be a part of–modern art.
Generative AI has made the line between influence and infringement thinner than ever. The doctrines of substantial similarity and fair use were designed to assess human judgment, not algorithmic mimicry. Still, the principle behind them persists: real creativity should transform rather than repeat. If the law can adapt to that truth—treating AI as a collaborator instead of a competitor—then technology and art can grow together under a framework that honors both innovation and artistic expression.
Edited by Ava Betanco-Born
Endnotes
[1] Copyright Act of 1790, 1 Stat. 124 (1790).
[2] Copyright Act of 1976, Pub. L. No. 94-553, 90 Stat. 2541 (1976).
[3] Nichols v. Universal Pictures Corp., 45 F.2d 119 (2d Cir. 1930).
[4] Thomson Reuters Enterprise Centre GmbH v. Ross Intelligence Inc., No. 20-cv-613-LPS (D. Del. 2025).