AI Music Copyright Lawsuits Reshape Licensing and Litigation in 2026

Two years after the major record labels filed the first coordinated copyright suits against AI music generators, the industry has arrived at a strange destination: the lawsuits everyone expected to settle the question of AI training have instead multiplied. Record labels cut licensing deals with the very companies they sued, a musicians' union is now suing those same labels, and a flurry of fresh complaints from publishers and indie artists keeps the courthouse doors swinging. The through-line of 2026 is that copyright infringement litigation has become a business model in itself — and the lawyers, armed with their own AI tools, have never been busier.

How the Suno and Udio Settlements Split the Majors

The June 2024 RIAA-coordinated suits against Suno and Udio have partially unwound, but not evenly. Universal Music Group settled with Udio in October 2025, establishing what observers describe as the first real licensing template for AI music generation10. Warner Music Group followed in November, settling with both Udio and Suno, with the Suno deal announced November 25 and enabling the startup to launch licensed AI models3. The Warner-Suno settlement came with an unusual kicker: Suno acquired Warner's Songkick ticketing platform as part of the arrangement1.

Sony Music Entertainment is the outlier. It has settled with neither platform and continues pushing both cases toward trial, making it the last major label standing in court25. Universal and Sony continue litigating against Suno in Massachusetts, with a protective order filed as recently as December 20254[9. Court schedules, according to one tracker, now push any US fair-use ruling in the music cases into 20274.

The settlements were not clean surrenders. Warner and Universal dropped their suits in exchange for licensing deals and jointly developed AI music platforms13, and reporting on the Universal-Udio terms describes per-generation royalties in the range of $0.002 to $0.0054. Forbes contributor Virginie Berger captured the cynicism many artists feel, arguing the sequence — launch on infringing data, then settle — effectively "made copyright infringement profitable" for the AI companies7.

The Discovery That Could Decide Suno's Fate

For the litigating holdouts, the evidence phase has been revelatory. Audio fingerprinting conducted in discovery against Suno identified millions of copyrighted recordings in its training data, a finding that now underpins a summary-judgment posture in the Sony case1. Suno, for its part, has refused to settle and is fighting on fair-use grounds, citing the Second Circuit's Bartz decision as support10.

The divergence across the majors is strategic, not accidental. Sony and Warner — at least in the Udio case — appear to be holding out for either better terms or a courtroom win that gives them leverage for higher royalties10. Independent artists filed separate class actions against both Suno and Udio in June 2025, ensuring that even a full settlement sweep by the majors would not clear the field410.

Everyone Suing Everyone: The Litigation Stack

The most striking development of 2026 is that the plaintiffs have themselves become defendants. The American Federation of Musicians sued Universal and Warner in mid-2026, alleging that by settling with Suno and Udio, the labels "are allowing those same AI companies to use the work of AFM-represented musicians" to train models on ingested recordings — without any artist share of the settlement money614. By September, Universal and Warner were asking a court to dismiss the union's claims12, and the Los Angeles Times has been probing who actually benefits when labels sign AI licensing deals11.

The litigation stack goes further. Top music publishers filed a new suit against an AI company in August, describing "blatant theft" — notably against Anthropic, fresh off its own $1.5 billion book-settlement, over pirated material allegedly used for training1720. Universal separately sued a music distributor it accuses of flooding platforms with "AI slop"16. And Suno has launched new models built on a major label catalog, a signal that the licensed-AI future the settlements promised is arriving even as the court fights continue15.

The bigger picture: trackers now catalog on the order of 200 active and resolved AI cases against 68 AI companies, spanning copyright, likeness, privacy, antitrust and employment claims in the US, EU and UK29. In Europe, a Munich court ruled in November 2025 that OpenAI's training on German song lyrics violated German copyright law — one of the first major European rulings against an AI developer, now on appeal29.

The Training-Data Rulings That Shaped the Landscape

The music fights play out against a broader body of training-data case law that has moved fast in two years. In June 2025, two Northern District of California decisions landed within days of each other. In Bartz v. Anthropic, Judge William Alsup held that training on copyrighted books was fair use — "transformative, spectacularly so" — but that storing pirated copies was not222325. In Kadrey v. Meta, Judge Vince Chhabria ruled for Meta on the authors' training claims while pointedly limiting his holding, noting the plaintiffs were "clear losers" on the record presented rather than on any general principle that AI training is always lawful2224.

The practical consequence of Bartz was a $1.5 billion class settlement by Anthropic, with final approval in July 2026 and an estimated payout of roughly $3,000 per work222321. That a fair-use win on training still produced a nine-figure payout — because of the pirated-acquisition theory — is the single most consequential signal in the current landscape: the liability fault line has shifted from training itself to how the data was obtained2225. Kadrey's remaining torrenting claims against Meta are still active2223, and major publishers, joined by author Scott Turow, have filed a new suit against Meta focused squarely on unlawful sourcing and demonstrable market harm — the two doctrines the earlier rulings left open30. Thomson Reuters v. Ross, an early decision finding against a fair-use defense for a competing legal-research tool, remains on appeal at the Third Circuit2223.

The music cases now sit on this fault line. Suno's fingerprinting revelation speaks not to training methodology but to acquisition of millions of recordings without authorization — precisely the theory that cost Anthropic $1.5 billion. Sony's bet on litigation looks less like stubbornness and more like a read of the case law.

Legal Tech's Inflection Point

All of this litigation is colliding with a legal profession that has itself crossed an adoption threshold. Generative AI use among legal professionals more than doubled between 2025 and 2026, from 31% to 69%, with 42% now using tools built specifically for legal practice, according to an industry survey of over 1,300 practitioners31. But institutions lag individuals: 54% of firms provide no AI training and 43% lack any formal policy, a readiness gap that analysts describe as the defining challenge of legal tech in 202631.

The stakes of that gap are visible in the cases themselves. The Suno discovery hinged on audio fingerprinting of training corpora — forensic tooling that barely existed in this context when the suits were filed1. Meanwhile, the Copyright Office's May 2025 pre-publication analysis of licensing, fair use and market questions circulates as a buyer-risk reference, a tacit acknowledgment that organizations procuring AI systems now need legal guidance on training-data provenance as a matter of course21. When the tracker literature is advising "buyers" on which court documents support which risk conclusions, AI copyright has stopped being a niche practice and become procurement diligence2122.

What the Coverage Agrees On — and Where It Diverges

Across sources, the core chronology is undisputed: UMG settled with Udio in October 2025, Warner settled with both platforms in November 2025, Sony has settled with nothing, and independent class actions continue2345. Where reporting diverges is on the trial horizon. One tracker puts a Udio trial in late 202610; another says US fair-use rulings in the music cases slip to 20274. And the framings differ sharply: the settlement news reads as industry normalization, while artist-adjacent coverage reads it as a double betrayal — labels monetizing catalogs through the same AI training they denounced in court76.

The reading I'd commit to: the settlements were not a resolution but a partition. Warner and Universal sold certainty; Sony is buying precedent. The Bartz line — training may be fair use, but stealing the corpus is not — has given rightsholders a durable theory that no settlement can erase, because it attaches to acquisition, not to output. That is why publishers keep filing, why the AFM is suing its employers, and why European courts are reaching opposite conclusions on the same conduct under different statutes. The 2026 reckoning isn't everyone suing everyone by accident; it's what happens when the law finally prices something the industry had been taking for free — and every party, including the lawyers' own AI tools, wants a share.