Eight Music Groups Just Told a Court AI Lyrics Can Be Illegal Without Copying a Single Line—Here’s the “Market Substitution” Test That Could Rewrite Fan Covers, Parodies, and Your Next Playlist
Eight major music organizations are pushing a judge to treat “market substitution” as the decisive fair-use issue in Concord v. Anthropic. If that framing wins, AI training and lyric outputs may require licensing—even without verbatim copying.
By TheMurrow Editorial
April 6, 2026
Key Points
- 1Eight music groups filed an amicus brief urging a judge to reject Anthropic’s fair-use defense in Concord v. Anthropic.
- 2Center the fight on “market substitution”: if Claude can deliver lyrics on demand, it may replace licensed lyric markets and trigger liability.
- 3Watch the 499-works case and the 20,000+ songs, $3B suit—early rulings could force licensing, output guardrails, and audits.
The music industry is done waiting for courts to “figure out” generative AI on their own time.
On March 30, 2026, a coalition of eight music organizations filed an amicus brief in federal court urging a judge to reject Anthropic’s fair-use defense in the lyrics lawsuit commonly referred to as Concord v. Anthropic. The brief backs music publishers seeking a partial summary judgment—an early ruling, before trial—that Anthropic’s use of lyrics is infringement, not fair use.
The filing is notable not just for who signed on—groups spanning major labels, independents, royalties, and artist advocacy—but for what they’re asking the court to treat as the decisive issue: market substitution. In plain English: if Claude can deliver lyrics on demand, and if AI companies can train on lyrics without paying, then the argument goes, AI is stepping into markets that copyright law is supposed to protect.
“The question isn’t whether AI is impressive. The question is whether it gets to replace a licensed market for songs and lyrics without paying for the privilege.”
— — TheMurrow (Pullquote)
Below is what happened, what the eight groups told the court, and why this fight over “substitution” may shape the future economics of music—and the rules under which AI products are built.
The amicus brief: eight groups, one message to the court
The news peg is straightforward: eight music-industry organizations filed an amicus brief supporting music publishers in their ongoing lawsuit against Anthropic, the AI company behind Claude. The brief urges the court to reject Anthropic’s claim that copying lyrics for training (and sometimes outputting lyrics) qualifies as fair use.
Reporting identifies the amici as:
- RIAA (Recording Industry Association of America)
- NMPA (National Music Publishers’ Association)
- A2IM (American Association of Independent Music)
- SoundExchange
- SONA (Songwriters of North America)
- BMAC (Black Music Action Coalition)
- Music Artists Coalition
- Artist Rights Alliance
The brief was filed Monday, March 30, 2026, according to Music Business Worldwide, which reported the development on April 1, 2026. The timing matters. The amici are backing the publishers’ motion for partial summary judgment in the earlier version of the case, sometimes described in coverage as the “original” lawsuit.
Reporting identifies the amici as:
- RIAA (Recording Industry Association of America)
- NMPA (National Music Publishers’ Association)
- A2IM (American Association of Independent Music)
- SoundExchange
- SONA (Songwriters of North America)
- BMAC (Black Music Action Coalition)
- Music Artists Coalition
- Artist Rights Alliance
The brief was filed Monday, March 30, 2026, according to Music Business Worldwide, which reported the development on April 1, 2026. The timing matters. The amici are backing the publishers’ motion for partial summary judgment in the earlier version of the case, sometimes described in coverage as the “original” lawsuit.
Who’s suing, and what’s being claimed
The underlying plaintiffs named in reporting are Universal Music Publishing Group, Concord Music Group, and ABKCO. Their core claim, as framed in coverage: Anthropic copied lyrics to train Claude, and Claude sometimes outputs lyrics, which the publishers say amounts to infringement—not fair use.
The organizations stepping in are not the plaintiffs. Their argument is broader: even if the court focuses on doctrine, the ruling will set expectations for how licensing markets work when AI systems ingest copyrighted text and can reproduce it in recognizable form.
The organizations stepping in are not the plaintiffs. Their argument is broader: even if the court focuses on doctrine, the ruling will set expectations for how licensing markets work when AI systems ingest copyrighted text and can reproduce it in recognizable form.
“Amicus briefs are rarely about a single defendant. They’re about the rules everyone will live under.”
— — TheMurrow (Pullquote)
Two lawsuits, two scales: why the numbers are getting bigger
Public coverage describes effectively two related disputes involving Anthropic and music publishers, and the difference in scale is part of the story.
The “original” case: hundreds of works, an early test of legal theories
The earlier publishers’ case is described as first brought in October 2023. Music Business Worldwide reports it involves 499 works. Even at that scale, the allegation is not a quirky edge case. It’s an assertion that an AI system trained on copyrighted lyrics—and capable of returning lyrics—crosses a legal line.
A motion for partial summary judgment is an attempt to secure a ruling on key issues without waiting for a full trial. For readers, that procedural detail matters: a court opinion at this stage can become a roadmap for other lawsuits, other industries, and other judges looking for a way through the thicket.
A motion for partial summary judgment is an attempt to secure a ruling on key issues without waiting for a full trial. For readers, that procedural detail matters: a court opinion at this stage can become a roadmap for other lawsuits, other industries, and other judges looking for a way through the thicket.
The later lawsuit: 20,000+ songs and a $3 billion demand
Coverage also describes a later and larger lawsuit reported as filed January 2026, covering more than 20,000 songs and seeking over $3 billion in statutory damages (as reported in TechCrunch and Music Business Worldwide coverage summarized by MBW).
The escalation in works and dollars signals a strategic choice. If the first case is about establishing liability and doctrine, the second is about leverage—how expensive it could become to train and operate large models if courts view unlicensed copying as infringement.
The escalation in works and dollars signals a strategic choice. If the first case is about establishing liability and doctrine, the second is about leverage—how expensive it could become to train and operate large models if courts view unlicensed copying as infringement.
Four key statistics readers should keep in mind
- 8 music groups filed the amicus brief (a wide coalition across industry segments).
- March 30, 2026 filing date for the brief (with April 1 coverage).
- 499 works alleged in the “original” case (per MBW).
- 20,000+ songs and $3+ billion in claimed statutory damages in the later suit (per coverage summarized by MBW).
Numbers don’t decide cases by themselves. Numbers do, however, shape what settlements look like—and how quickly an industry decides it must negotiate rather than litigate.
- March 30, 2026 filing date for the brief (with April 1 coverage).
- 499 works alleged in the “original” case (per MBW).
- 20,000+ songs and $3+ billion in claimed statutory damages in the later suit (per coverage summarized by MBW).
Numbers don’t decide cases by themselves. Numbers do, however, shape what settlements look like—and how quickly an industry decides it must negotiate rather than litigate.
8
Music-industry organizations joined the amicus brief, spanning labels, publishers, indies, royalties, and artist-advocacy groups.
March 30, 2026
Filing date of the amicus brief supporting the publishers’ partial summary judgment push (reported April 1, 2026).
499 works
Scale of the “original” case (first brought October 2023), as reported by Music Business Worldwide.
20,000+ songs / $3+ billion
Scale and damages demand in the later lawsuit reported as filed January 2026 (per coverage summarized by MBW).
The heart of the argument: what “market substitution” means in fair use
Fair use law can sound abstract until you reach the fourth factor: the effect of the use upon the potential market for or value of the copyrighted work. That’s where substitution lives.
The amici are pressing the court to treat the fourth factor as central. Music Business Worldwide’s reporting explicitly frames their argument as focused on substitutional harm—the idea that AI outputs can displace existing or likely licensing markets for lyrics.
The amici are pressing the court to treat the fourth factor as central. Music Business Worldwide’s reporting explicitly frames their argument as focused on substitutional harm—the idea that AI outputs can displace existing or likely licensing markets for lyrics.
The doctrinal baseline: the market is part of the property
Courts evaluating fair use traditionally ask whether the challenged use undermines:
- the market for the original work, and
- traditional, reasonable, or likely-to-be-developed markets for licensing
That second clause is the flashpoint. It allows copyright holders to argue that even if an AI tool doesn’t reduce sales of a particular song today, it may erode licensing markets that are both foreseeable and legitimate—lyrics databases, authorized displays, and other uses where permission and payment have historically mattered.
The amici’s position, as described in coverage, aims to bring AI training and AI outputs into that same frame: if an AI product can serve the same consumer purpose as a licensed lyrics offering, then the harm is not hypothetical.
- the market for the original work, and
- traditional, reasonable, or likely-to-be-developed markets for licensing
That second clause is the flashpoint. It allows copyright holders to argue that even if an AI tool doesn’t reduce sales of a particular song today, it may erode licensing markets that are both foreseeable and legitimate—lyrics databases, authorized displays, and other uses where permission and payment have historically mattered.
The amici’s position, as described in coverage, aims to bring AI training and AI outputs into that same frame: if an AI product can serve the same consumer purpose as a licensed lyrics offering, then the harm is not hypothetical.
How substitution differs from “being inspired by”
The conversation often collapses into a muddle: training is “learning,” outputs are “new,” and the system is “transformative.” Fair use doctrine can accommodate transformation. It is less forgiving when transformation becomes a substitute.
A practical way to understand the dispute is to separate two questions:
1) Did Anthropic copy protected material to build a product?
2) Does the product function as a replacement for a market that copyright law recognizes?
The publishers and amici, as reported, want the judge to answer yes to both—then treat that as fatal to fair use.
A practical way to understand the dispute is to separate two questions:
1) Did Anthropic copy protected material to build a product?
2) Does the product function as a replacement for a market that copyright law recognizes?
The publishers and amici, as reported, want the judge to answer yes to both—then treat that as fatal to fair use.
“Fair use tolerates transformation. It balks at replacement.”
— — TheMurrow (Pullquote)
Key Insight: The “Market Substitution” Lens
The amici want the judge to treat fair use factor four—market effect—as the decisive test: if AI replaces licensed lyric access, it’s not fair use.
What the music groups told the court—without pretending it’s only about lyrics
The amici brief, as described by Music Business Worldwide, attacks the premise that unlicensed copying is excusable merely because it fuels AI innovation. The brief supports the publishers’ motion and urges the court to reject Anthropic’s fair-use defense.
Even without quoting the filing line-by-line, the contours are clear from the reporting: the music groups argue that copying lyrics without permission is infringement, and the presence of AI does not grant a special exemption.
Even without quoting the filing line-by-line, the contours are clear from the reporting: the music groups argue that copying lyrics without permission is infringement, and the presence of AI does not grant a special exemption.
“Unlicensed copying” as the central fact, not a technical detail
The amici’s framing, per MBW, treats the lack of licensing as a moral and economic problem—not just a procedural oversight. That matters because AI cases often drift into technicalities: training corpora, tokenization, weights, and whether the model stores verbatim text.
The brief appears to insist on an older, simpler idea: when a business uses copyrighted works to build a commercial product, the default rule is permission and payment—especially where the product can output protected material.
The brief appears to insist on an older, simpler idea: when a business uses copyrighted works to build a commercial product, the default rule is permission and payment—especially where the product can output protected material.
A coalition designed to look like the market itself
The list of amici reads like a map of music’s rights ecosystem:
- Publishers and labels (through organizations like NMPA and RIAA)
- Independents (A2IM)
- Royalties and neighboring rights (SoundExchange)
- Songwriters and artist advocacy groups (SONA, Artist Rights Alliance, Music Artists Coalition)
- Equity and representation (BMAC)
That breadth strengthens the substitution argument. The amici aren’t merely saying “we dislike AI.” They’re saying: many parts of the music economy rely on licensing norms, and unlicensed AI training destabilizes them across the board.
- Publishers and labels (through organizations like NMPA and RIAA)
- Independents (A2IM)
- Royalties and neighboring rights (SoundExchange)
- Songwriters and artist advocacy groups (SONA, Artist Rights Alliance, Music Artists Coalition)
- Equity and representation (BMAC)
That breadth strengthens the substitution argument. The amici aren’t merely saying “we dislike AI.” They’re saying: many parts of the music economy rely on licensing norms, and unlicensed AI training destabilizes them across the board.
A procedural moment with outsized impact
Because the amicus supports a partial summary judgment motion, the aim is not just to win eventually. The aim is to define the rules now, in a way that influences negotiations and follow-on litigation.
A court that agrees with the publishers on fair use could narrow the range of defenses available to AI companies in other copyright disputes—inside and outside music.
A court that agrees with the publishers on fair use could narrow the range of defenses available to AI companies in other copyright disputes—inside and outside music.
Editor’s Note
This filing is positioned as rule-setting: an early ruling on fair use could become a roadmap for other AI copyright fights.
Anthropic’s fair-use defense: the strongest version of the opposing view
A fair reading of the controversy requires acknowledging what fair use defenses generally try to establish, especially in AI cases.
Although the research here focuses on the amici and publishers’ framing (and does not provide Anthropic’s detailed legal arguments), the standard fair-use posture in generative AI disputes tends to emphasize some combination of: transformation, technological purpose, and distance from the original market.
Although the research here focuses on the amici and publishers’ framing (and does not provide Anthropic’s detailed legal arguments), the standard fair-use posture in generative AI disputes tends to emphasize some combination of: transformation, technological purpose, and distance from the original market.
The argument AI companies want courts to accept
In broad terms, AI companies typically resist the “substitution” framing by arguing that:
- training is not the same as republishing,
- outputs are not necessarily copies, and
- the system’s purpose differs from the purpose of the original work
From that perspective, the fourth factor becomes a fight over what counts as a legitimate “market.” Are publishers entitled to a licensing market for training data? Are they entitled to control machine learning uses that don’t present full songs, but can reproduce lyrics in response to prompts?
The amici brief is designed to push the court toward a stricter answer: yes, licensing markets count, and yes, outputs that can deliver lyrics threaten those markets.
- training is not the same as republishing,
- outputs are not necessarily copies, and
- the system’s purpose differs from the purpose of the original work
From that perspective, the fourth factor becomes a fight over what counts as a legitimate “market.” Are publishers entitled to a licensing market for training data? Are they entitled to control machine learning uses that don’t present full songs, but can reproduce lyrics in response to prompts?
The amici brief is designed to push the court toward a stricter answer: yes, licensing markets count, and yes, outputs that can deliver lyrics threaten those markets.
Why the “sometimes outputs lyrics” detail is a big deal
Music Business Worldwide’s reporting notes that Claude sometimes outputs lyrics. That detail matters because it ties training to a user-facing act that looks like ordinary infringement: providing copyrighted text to users.
A system that merely “learns patterns” is harder to evaluate. A system that can return lyrics on request is easier for a judge—or a jury—to understand as a substitute for authorized access.
If you can get lyrics from a chatbot, you may not need to visit a licensed lyrics service, buy a digital booklet, or rely on authorized reproductions. That’s the substitution theory in its most intuitive form.
A system that merely “learns patterns” is harder to evaluate. A system that can return lyrics on request is easier for a judge—or a jury—to understand as a substitute for authorized access.
If you can get lyrics from a chatbot, you may not need to visit a licensed lyrics service, buy a digital booklet, or rely on authorized reproductions. That’s the substitution theory in its most intuitive form.
Real-world implications: what this means for artists, fans, and AI products
Court battles can feel remote until they reshape the tools people use daily. The practical stakes here fall into three buckets: artist income, consumer access, and product design.
For artists and songwriters: licensing norms are the income floor
Organizations like SONA and artist coalitions exist because creators’ income often depends on licensing structures that feel invisible to listeners. If courts accept that training and output can happen without licensing, creators fear a race to the bottom: a system where the largest platforms get the most valuable catalog for free, while everyone else is expected to monetize “exposure.”
The amici’s intervention suggests an insistence on continuity: new technology, old rule—permission and payment.
The amici’s intervention suggests an insistence on continuity: new technology, old rule—permission and payment.
For fans: convenience versus provenance
Many users don’t care where lyrics come from; they care whether the answer is correct and fast. AI systems thrive on that preference.
A ruling that treats lyrics outputs as infringing could lead AI products to add friction:
- refusals to provide lyrics verbatim,
- licensing deals with publishers,
- tighter content filters that reduce false positives but also block legitimate queries
Readers should expect tradeoffs. A more rights-respecting ecosystem may also be a less seamless one.
A ruling that treats lyrics outputs as infringing could lead AI products to add friction:
- refusals to provide lyrics verbatim,
- licensing deals with publishers,
- tighter content filters that reduce false positives but also block legitimate queries
Readers should expect tradeoffs. A more rights-respecting ecosystem may also be a less seamless one.
For AI companies: design choices become legal choices
If “market substitution” becomes the decisive lens, product teams can’t treat legal as a cleanup crew. Design decisions—what the model is trained on, what it can output, how it responds to prompts for lyrics—become front-line risk management.
A world in which 499 works become the foundation of a summary judgment ruling is also a world where the 20,000+ song case becomes an existential threat. Even the reported $3+ billion demand functions as a warning sign to the market: ignoring licensing may be expensive enough to change behavior.
A world in which 499 works become the foundation of a summary judgment ruling is also a world where the 20,000+ song case becomes an existential threat. Even the reported $3+ billion demand functions as a warning sign to the market: ignoring licensing may be expensive enough to change behavior.
Practical takeaways: how to read the next phase of the case
Legal coverage can become a fog of filings. Here are concrete signals worth watching as this proceeds.
What to watch in court filings and rulings
- How the judge frames the fourth factor (market effect): A strong statement on substitution can influence other AI copyright cases.
- Whether the court treats lyrics licensing as a “likely-to-be-developed” market for AI training: That language is pivotal in fair use doctrine.
- How much weight the court gives to “sometimes outputs lyrics”: If outputs are central, AI companies may prioritize preventing verbatim reproduction.
- Whether the court treats lyrics licensing as a “likely-to-be-developed” market for AI training: That language is pivotal in fair use doctrine.
- How much weight the court gives to “sometimes outputs lyrics”: If outputs are central, AI companies may prioritize preventing verbatim reproduction.
What a settlement would probably require
No one should assume a settlement is imminent. Still, the shape of any deal is predictable in broad strokes:
- a licensing framework for lyrics used in training,
- guardrails on outputs, and
- reporting/auditing obligations that convince rights holders the bargain is real
The amici brief is part legal argument, part negotiation posture. It tells the court—and Anthropic—that the industry wants rules that look like licensing, not forgiveness.
- a licensing framework for lyrics used in training,
- guardrails on outputs, and
- reporting/auditing obligations that convince rights holders the bargain is real
The amici brief is part legal argument, part negotiation posture. It tells the court—and Anthropic—that the industry wants rules that look like licensing, not forgiveness.
How this could affect other creative industries
Even though the dispute is about music lyrics, the logic of substitution is portable. If a court accepts that AI outputs can replace licensed access, other rights holders—from book publishers to news organizations—will cite that reasoning.
A narrow case about lyrics can become a wide precedent about markets.
A narrow case about lyrics can become a wide precedent about markets.
Signals worth watching next
- ✓Judge’s language on fair use factor four (market effect)
- ✓Treatment of “likely-to-be-developed” licensing markets for AI training
- ✓Weight given to Claude “sometimes outputs lyrics” behavior
- ✓Whether parties move toward licensing + output guardrails + auditability
A ruling about value, not novelty
The amicus brief filed on March 30, 2026 reflects an industry trying to pin a simple principle to a complex technology: creative work has an owner, and markets built on that work deserve protection against free substitutes.
Music organizations are not asking the court to ban AI. They’re asking the court to treat AI companies like every other commercial actor that benefits from copyrighted expression: get permission, or face liability.
If the court buys the substitution framing, the impact will extend beyond Anthropic and beyond lyrics. It will tell creators whether copyright still functions as a market-making tool in the age of generative systems—or whether “training” becomes a carveout that swallows the rule.
The outcome won’t settle society’s debate about AI. It will settle something more concrete: who gets paid, and who gets to build on whose work.
Music organizations are not asking the court to ban AI. They’re asking the court to treat AI companies like every other commercial actor that benefits from copyrighted expression: get permission, or face liability.
If the court buys the substitution framing, the impact will extend beyond Anthropic and beyond lyrics. It will tell creators whether copyright still functions as a market-making tool in the age of generative systems—or whether “training” becomes a carveout that swallows the rule.
The outcome won’t settle society’s debate about AI. It will settle something more concrete: who gets paid, and who gets to build on whose work.
T
About the Author
TheMurrow Editorial is a writer for TheMurrow covering entertainment.
Frequently Asked Questions
What is “Concord v. Anthropic” about?
Public coverage describes a lawsuit brought by music publishers—including Universal Music Publishing Group, Concord Music Group, and ABKCO—alleging Anthropic copied song lyrics to train Claude and that Claude sometimes outputs lyrics. The publishers argue this is copyright infringement and not protected by fair use.
Which eight music groups filed the amicus brief?
Reporting identifies the amici as RIAA, NMPA, A2IM, SoundExchange, SONA, BMAC, Music Artists Coalition, and Artist Rights Alliance. They filed an amicus brief supporting the publishers and urging the court to reject Anthropic’s fair-use defense.
When was the amicus brief filed, and what stage is the case in?
Music Business Worldwide reports the brief was filed March 30, 2026. The amici brief supports the publishers’ motion for partial summary judgment in the earlier (“original”) case—meaning the publishers seek a court ruling on key issues before trial.
What does “market substitution” mean in this context?
“Market substitution” refers to the fair use fourth factor: whether the challenged use harms the market for the original work or its licensing markets. Here, the argument is that AI training and outputs can replace licensed ways people access lyrics, undermining existing or reasonably expected markets.
Why are there two different Anthropic lawsuits mentioned in coverage?
Coverage describes an earlier case (first brought October 2023) involving 499 works, and a later lawsuit reported as filed January 2026 covering more than 20,000 songs and seeking over $3 billion in statutory damages. They are related in subject matter but differ greatly in scale.
Why does it matter that Claude “sometimes outputs lyrics”?
That detail, highlighted in reporting, connects training data disputes to user-facing behavior. If a chatbot can deliver lyrics on demand, rights holders argue it can function as a substitute for authorized lyrics sources—strengthening the claim of market harm under the fourth fair-use factor.
