Perspectives· AI Copyright
AI Lyrics Copyright Hearing Ends Without a Ruling, and Reviews Disagree on Who's Ahead
Songwriters, music publishers and AI companies are waiting for a decision that could shape other AI copyright cases. A close look at the hearing shows the question is still open, and even careful readings of the arguments disagree about which side the record currently favors.
AI4E Perspectives · Published 2 November 2026 · Sources as of 2 November 2026
Compares answers from
- GPT-6 Astra
- Claude Sonnet 5
- Gemini 4 Argon
- Grok 4.7
- Mistral Large
- DeepSeek V4 Pro
A federal judge heard arguments over whether Anthropic's AI training on song lyrics counts as fair use, but she did not rule at the hearing. She questioned the publishers' claim that Claude's lyric-writing threatens their market, yet the record leaves open how much that skepticism should count. Reviews of the hearing disagree sharply on whether the reported evidence currently favors Anthropic or neither side, a disagreement that matters because the case could influence other pending AI copyright suits.
An Unresolved Hearing
At the hearing, Anthropic's lawyer argued that training on lyrics produces transformative content protected by fair use, and said publishers could not point to specific instances of market harm. The publishers' lawyer argued Claude generated competing content and cited nearly 120,000 requests for synthetic lyrics over six months as a sign the market could be overwhelmed. The judge questioned the publishers' claim that this threatens their market, but she did not rule.
Because no ruling was issued, every characterization of which argument came out ahead at the hearing is a reading of reported arguments and judicial questioning, not a legal finding.
Evidence: Agreement 1 · Agreement 2 · Agreement 3 · Agreement 4
Disagreement Over Who the Evidence Favors
Reviews of the hearing split on this point. Some concluded Anthropic's position looks better supported, pointing to the publishers' lack of concrete harm data and the judge's questioning. Others declined to say either side is ahead, noting that the reported questioning toward the publishers may reflect which arguments the reporting highlighted, rather than the complete hearing record.
Separately, there is no agreement on whether an absence of proven market harm should carry more legal weight for a transformative-use defense, or whether requiring such proof structurally favors large AI companies. The source material describes the arguments made but not the governing legal standard.
Evidence: Difference 2 · Difference 1
What Two Earlier Rulings Show
Two other judges handling similar AI-training fair-use questions reached different outcomes: one sided with Anthropic on training using books, and another ruled for Meta while cautioning that such training would not always qualify as fair use. Both earlier cases ended in rulings for the AI companies, so the divide is less about who won than about how far fair-use protection extends, a distinction that complicates treating either ruling as a template for this lyrics case.
Evidence: Agreement 5 · Difference 3
A Compatible but Incomplete Quote
Anthropic's lawyer cited a UMG statement describing AI as a net positive for the music industry, offered as a counter to the publishers' harm claims. Reviewers generally agreed that a broad statement about AI's industry-wide benefit does not rule out a specific infringement claim over unlicensed lyrics; the two things can be true at once. But the source material does not include the original statement's setting, so how broadly UMG intended it remains unclear.
Evidence: Agreement 7 · Assumption 4
Interpretation
AI4E assessment
This is AI4E's interpretation of the reviewed material, not an established fact. The record supports describing an unresolved hearing with competing arguments, but it does not support declaring either side clearly ahead.
- Supported by the comparison
- The hearing ended without a ruling, and the judge questioned the publishers' market-harm claim without deciding the case.
- A general statement that AI benefits the music industry is logically compatible with a specific infringement claim over unlicensed lyrics.
- The split between two earlier AI-training rulings means neither can be treated as a reliable template for how this lyrics case will turn out.
- Premature to conclude
- Concluding that the reported evidence clearly favors Anthropic's fair-use defense.
- Treating the absence of proven market harm as evidence of an established legal rule that favors large AI companies.
- What would change this view
- A written ruling from the judge handling this case.
- Concrete data on lost songwriting income or licensing deals tied to AI-generated lyrics.
- The full context of UMG's net-positive statement.
Still unresolved
- Whether training on song lyrics counts as fair use in this case.
- Whether the publishers can show concrete market harm from AI-generated lyrics beyond request volume.
- How a ruling in this case would affect other pending suits by authors, news outlets and visual artists.
- What UMG's net-positive statement originally meant and in what setting it was made.
Model statements are not verified facts, and agreement between models is not independent confirmation. Inspect the comparison, original answers and dated sources.