The plaintiff is not the artists
If Phil Rosen succeeds, the individual defendants could be ordered to pay damages to Alphabet, return compensation, and accept court-ordered changes to the company's oversight of copyright and biometric-privacy compliance. The complaint does not specify a damages amount. Alphabet, rather than Rosen or the musicians whose works are at issue, would receive any damages awarded.
This is not a claim for damages brought by musicians or voice performers. Rosen is pursuing alleged corporate harm on Alphabet's behalf. His claim is fiduciary: that the board adopted what the complaint calls an unlawful business strategy — building AI on copyrighted works and people's voices instead of obtaining licenses first — and then misled shareholders about the oversight in place.
The music claim, in numbers
The complaint's most specific passage concerns MusicLM, the music generator Google introduced in January 2023. It says the model was trained on five million audio clips totaling 280,000 hours of music, and that Google did not identify the specific recordings, rights holders or licenses in that corpus. It alleges the wider Lyria family — Lyria, Lyria 2, Lyria 3 and a system called Mulan — was built the same way: audio extracted from YouTube recordings, converted into machine-readable form, cut into clips, with the information identifying artists and their copyright notices stripped out. It notes that four researchers Google named as core Lyria contributors later left to launch Udio, which the RIAA sued for the major labels in June 2024.
Read those numbers as one side's filing, because that is what they are. Much of this music section draws on a proposed class action that independent musicians filed against Google in the same Chicago court in March, and some passages repeat that lawsuit's wording. Google has asked for that case to be thrown out, arguing that by accepting YouTube's terms of service the musicians granted a license extending to YouTube's affiliates, Google included. No court has ruled that the training was unlawful, and the primary report of Rosen's filing carries no response from Google.
The governance claim is the new part
What Rosen adds is the paper trail aimed at the boardroom. Alphabet's 2024 proxy told shareholders its board received regular updates from managers immersed in AI Principles designed to keep development safe and trustworthy; the complaint says those statements disappeared from the 2025 and 2026 proxies, and calls the originals materially false and misleading. The complaint also quotes Alphabet's 2025 proxy statement, which opposed a shareholder proposal for a report on AI data usage. In the passage reproduced in the filing, the board said Google operates "an authoritative, comprehensive, and auditable system of record for all of our AI assets," with "detailed tracking of data origins and usage (auditable lineage)." Rosen alleges that this assurance was misleading.
Then comes the figure that gives the fiduciary claim its weight. Between April 2023 and November 2025, Alphabet bought back 1.147 billion of its own shares for USD $169.6 billion. Rosen alleges those repurchases were made at prices artificially inflated by the undisclosed copyright and biometric exposure — the company, in effect, overpaying for its own stock while its directors withheld a liability they knew about. The filing also points out that one defendant, Chief Legal Officer Kent Walker, published a Google policy paper in June arguing that training on publicly available web data should be protected as fair use. Bloomberg Law reported that a shareholder named Phil Rosen filed a similar derivative suit against Apple's board in August. The reporting does not establish whether he is the same person who filed this Alphabet case. Taken together, the two filings suggest that derivative litigation may become another route for shareholders seeking to challenge how AI developers manage copyright exposure.
None of this establishes that Google's training was unlawful. It establishes something narrower and, for Alphabet's investors, harder to wave away: a shareholder is now asking a court to price the gap between what the board said about data oversight and what the complaint says the board knew. The copyright cases will address whether the underlying training practices infringed protected works. This derivative action asks whether Alphabet's directors and executives breached their duties in overseeing those practices and disclosing the resulting risks — and whether they should be held financially accountable to the company.
Sources
- [1] Music Business Worldwide, "Alphabet shareholder sues Sundar Pichai and 12 others over training Google's AI, including Lyria, on copyrighted works", published November 2, 2026, 18:32 UTC (primary report for this article; MBW obtained the full 79-page complaint — the derivative structure, the 13 defendants, the Lyria/MusicLM allegations, the 5 million clips / 280,000 hours figure, the BIPA voiceprint claim, the proxy-statement passages, the 1.147 billion shares / USD $169.6 billion buybacks, Pichai's compensation, the March class action and Google's dismissal argument, the Kent Walker paper, the Apple parallel suit, and MBW's correction of the complaint's Anthropic $1.5B attribution — read first-hand in full)Read source
- [2] Music Ally, "AI roundup: Lyria lawsuit, Anthropic's Q3, OpenAI pile-in", published November 3, 2026, 12:30 UTC (the roundup that recirculated the filing; confirms the derivative structure and the Lyria focus)Read source
- [3] Digital Music News, "Alphabet Directors Face Shareholder Suit Over AI Training Policies", published November 2, 2026, updated November 3, 2026, 04:54 UTC (corroborating report; the fiduciary-duty, unjust-enrichment and Exchange Act claims, the audit-committee allegation and the buyback-inflation claim)Read source