美国法院审理用歌词训练 AI 是否属合理使用，印度拟推强制许可

This case was filed by Universal Music Group (UMG), Concord and ABKCO in 2023, and they alleged that Anthropic used lyrics from hundreds of songs by artists such as Beyoncé, the Rolling Stones and the Beach Boys for training Claude. Both sides have sought summary judgment, asking the court to decide the case, or specific legal issues, without a full trial.

What did each side argue?

Anthropic, through its lawyer, Sonal Mehta of WilmerHale, argued that this usage would fall under fair use since the training leads to transformative new content. Further, on the point of market harm, she argued that AI does not reduce demand for song lyrics; in fact, UMG itself describes AI as a net positive for the industry.

The publishers, through their lawyer, Matt Oppenheim of Oppenheim + Zebrak, argued that this training harms songwriters as Anthropic uses the lyrics for generation of competing content. He said users made nearly 120,000 requests to Claude to generate ‘synthetic lyrics’ over six months, and that models like Claude “could overwhelm the market with AI-generated content very easily”.

Judge Lee pressed both sides on fair use and the claim that Claude has the ability to write lyrics which would threaten the publishers in the market, Reuters reported.

What have other US courts said?

Two other judges in the Northern District of California reached different conclusions in 2025 about whether AI training qualifies as fair use, Reuters noted. In 2025, now-retired Judge William Alsup had ruled that while pirated copies are not allowed for training, using lawfully acquired books to train AI models could qualify as fair use. Then, Judge Vince Chhabria, in a similar suit, ruled that AI training in many cases does not count as fair use.

Anthropic also faces lawsuits involving song lyrics brought by Sony Music Publishing and Round Hill Music.

Where does India stand?

What MediaNama has asked

Anthropic:

Whether Anthropic has licensed lyrics or music catalogues from any Indian rights holder to train Claude

Whether the safeguards Claude applies against reproducing lyrics cover songs in Indian languages

The publishers, through their counsel:

How the figure of nearly 120,000 synthetic-lyric requests was measured, and whether it forms part of the court record

Whether the publishers, or their Indian affiliates, intend to bring similar claims in India

IMI:

Whether AI developers have approached IMI or its members to license lyrics or recordings for training, and on what terms

How royalty rates under DPIIT’s proposed licence should account for AI outputs that compete with the original songs

MediaNama has sought comment from Anthropic, the publishers’ counsel and IMI. This story will be updated when they respond.

美国法院审理用歌词训练 AI 是否属合理使用，印度拟推强制许可

推荐理由

主题

标签

正文