Anthropic vs. Music Publishers: The Fair Use Fight Over AI Training

Seed story: "Anthropic, music publishers square off on 'fair use' defense in AI copyright case" (Reuters) · search original Written from facts verified across 2 news report(s) — original explainer, not a copy or translation. Sources listed at the end.

The outcome of the ongoing legal battle between Anthropic and major music publishers could fundamentally reshape how creators protect their intellectual property in the age of generative AI. With a federal judge recently questioning the validity of the "fair use" defense for training models on millions of song lyrics, the case poses a critical question for freelancers and artists: does the transformative nature of AI development outweigh the economic impact on the original creators?

The Latest Hearing in the Anthropic Copyright Dispute

Procedural Stalemate

On October 21, 2026, US District Judge Eumi Lee presided over a critical hearing in San Jose regarding the ongoing copyright dispute between Anthropic and major music publishers. Rather than issuing a definitive ruling, Judge Lee focused on questioning both parties to clarify the nuances of the fair use defense. This procedural pause leaves the case in a state of uncertainty, with no immediate resolution for the stakeholders involved.

The hearing highlighted the core legal friction points:

Anthropic’s attorney, Sonal Mehta, argued that training creates transformative content and that publishers failed to prove specific market harm.
Publishers’ attorney, Matt Oppenheim, countered that Claude received nearly 120,000 requests to generate synthetic lyrics over six months.
The case involves alleged infringement of hundreds of songs by artists like Beyonce and the Rolling Stones.

For creators, this lack of a final ruling means that contract terms regarding AI training data remain ambiguous. Until Judge Lee issues a decision, artists and publishers must navigate a legal landscape where the boundaries of permissible use are still being actively debated in the courtroom.

Why the Fair Use Defense Is Central to This Case

The Legal Battle Over Transformative Use

Anthropic’s legal team, led by attorney Sonal Mehta, contends that training the Claude chatbot on millions of song lyrics constitutes "transformative" use. They argue that this process creates entirely new content, a concept central to the fair use doctrine. Crucially, Mehta emphasized that the publishers failed to pinpoint specific instances where this training caused measurable market harm to the original works.

Conversely, the publishers, represented by Matt Oppenheim, argue that the scale of the impact is undeniable. They highlighted that Claude received nearly 120,000 requests to generate synthetic lyrics over just six months. This data suggests a direct substitution for human creativity, undermining the argument that the AI’s output is harmless or distinct from the source material.

Anthropic’s stance: Training is transformative and lacks specific market harm.
Publishers’ counter: High volume of synthetic lyrics proves market disruption.
Key evidence: 120,000 synthetic lyric requests in six months.

For creators, this debate determines whether AI developers can legally ingest copyrighted works without compensation, directly impacting future licensing negotiations and royalty structures.

The Scale of the Alleged Infringement

The core of the dispute centers on the sheer volume of copyrighted material allegedly used to train Claude. According to reports, the lawsuit filed by Universal Music Group, Concord, and ABKCO in 2023 alleges infringement across hundreds of songs. The complaint specifically names major artists, including Beyoncé, the Rolling Stones, and the Beach Boys, highlighting the breadth of the catalog involved.

Beyond the training data, publishers point to the output generated by the model. Their attorney, Matt Oppenheim, noted that Claude received nearly 120,000 requests to generate synthetic lyrics over a six-month period. This high volume of user-driven generation suggests a direct commercial application of the trained data.

Plaintiffs: Universal Music Group, Concord, and ABKCO
Named Artists: Beyoncé, the Rolling Stones, the Beach Boys
Alleged Infringement: Hundreds of songs
Synthetic Output: ~120,000 lyric requests in six months

These figures are critical for determining whether the use constitutes a market substitute, a key factor in fair use analysis.

Precedents from Recent AI Copyright Rulings

A Split in Judicial Interpretation

Recent federal rulings have created a fragmented legal landscape for AI developers. In 2025, US District Judge William Alsup ruled in favor of Anthropic, determining that training AI models on books constituted fair use. However, this decision was immediately complicated by a ruling from US District Judge Vince Chhabria just two days later. While Chhabria sided with Meta, he explicitly warned that AI training would not be considered fair use in many circumstances.

This divergence leaves creators and rights holders in a state of uncertainty. The core conflict lies in how judges weigh the "transformative" nature of training against potential market harm.

Judge Alsup’s ruling suggests broad protection for training on literary works.
Judge Chhabria’s warning implies a narrower, context-dependent standard.
The Anthropic case now hinges on whether these precedents apply to song lyrics.

For creators, this inconsistency means that contract terms regarding AI licensing remain critical. Without a unified judicial standard, relying solely on fair use defenses is risky, making explicit contractual permissions essential for protecting rights and ensuring fair compensation.

Implications for Creative Contracts and Rights

Contractual Shifts

The pending decision in this San Jose case will likely force a fundamental reevaluation of standard licensing agreements. If Judge Lee rejects the fair use defense, creators may find that implicit permissions are no longer sufficient, necessitating explicit, paid licenses for AI training. This shift could alter the definition of ownership, moving it from mere copyright retention to active data governance rights.

Key implications for creators include:

Explicit Consent: Contracts may require specific clauses permitting or prohibiting model training.
Revenue Sharing: New frameworks might mandate compensation for the use of lyrics in synthetic generation.
Opt-Out Mechanisms: Standardized methods for creators to exclude their work from datasets.

Ultimately, the outcome determines whether AI companies can rely on broad fair use arguments or must negotiate directly with rights holders, significantly impacting how creators monetize their intellectual property in the digital age.

Strategies for Creators in the AI Era

Protecting Your Work in the AI Era

As legal standards shift, independent professionals must proactively manage their intellectual property. With recent rulings like Judge Alsup’s 2025 decision favoring Anthropic, the landscape for training data is complex. Creators should not assume their work is automatically protected against AI ingestion. Instead, focus on clear contractual language.

To navigate this evolving terrain, consider these actionable steps:

Audit existing contracts for explicit AI licensing clauses.
Register copyrights to establish a clear timeline of ownership.
Monitor usage for unauthorized synthetic generation, similar to the 120,000 requests cited in the Anthropic case.

By defining your rights upfront, you ensure that any licensing opportunities are transparent. This approach helps secure fair compensation and prevents ambiguity in future disputes over how your creative output is utilized in machine learning models.

FAQ

What is the current status of the copyright lawsuit between Anthropic and music publishers?

As of October 21, 2026, US District Judge Eumi Lee has not issued a ruling on the case. The judge presided over a hearing in San Jose where both sides presented arguments regarding the fair use defense, but no final decision was made at that time.

How is Anthropic defending its use of song lyrics to train its Claude chatbot?

Anthropic argues that using millions of song lyrics to train its AI model constitutes fair use under US copyright law. Their attorney, Sonal Mehta, contends that the process creates transformative new content and that the publishers have failed to identify specific instances of market harm.

What evidence have music publishers presented to support their claim of copyright infringement?

Music publishers, including Universal Music Group, Concord, and ABKCO, allege that Anthropic infringed copyrights in hundreds of songs by artists such as Beyonce and the Rolling Stones. Their attorney Matt Oppenheim highlighted that Claude received nearly 120,000 requests to generate synthetic lyrics over a six-month period.