On October 30, Alphabet shareholder Phil Rosen sued Sundar Pichai, Larry Page, Sergey Brin and 10 other directors and executives in federal court in Chicago. He says they let Google train its AI on copyrighted books, images, voices and music. The music part covers the Lyria models. This Alphabet shareholder lawsuit is a derivative suit, so any damages would go to Alphabet itself. Below you’ll find the timeline, the music claims, and what it all means if you put songs on YouTube.
A derivative suit is a strange thing. A shareholder sues the company’s own leaders on the company’s behalf, claiming they hurt it. If the Alphabet shareholder lawsuit wins, the cheque goes from the bosses to Alphabet. So Google would get paid back for Google’s own AI training. That’s the business model.
The Alphabet shareholder lawsuit says the plan was to pay damages later
The complaint behind the Alphabet shareholder lawsuit, reported by Music Business Worldwide, accuses all 13 defendants of “an unlawful business strategy”. In Rosen’s telling, nobody forgot the licences. They skipped them on purpose. The plan, the complaint says, was “carefully crafted to avoid spending the time and money necessary to obtain lawful licenses or permission from the copyright holders, and instead to simply pay damages later.”
It calls this the “ask forgiveness not approval” model. Rosen says the leaders broke their fiduciary duties (the legal duty to act in the company’s interest). They built AI on what Bloomberg Law sums up as stolen data, then misled shareholders about it in annual filings.
Pichai, Page and Brin are joined by Alphabet Chair John L. Hennessy, six other directors, CFO Anat Ashkenazi, President and Chief Investment Officer Ruth M. Porat, and Kent Walker, Alphabet’s Chief Legal Officer. The complaint says Pichai alone got more than $30 million in pay from 2023 to 2025. Rosen wants some back. He also wants the court to force “effective oversight of compliance with copyright and BIPA laws”. BIPA is Illinois’ Biometric Information Privacy Act, which requires written consent before a company collects biometric data such as voiceprints. Your voice counts.
What happened so far
The story starts years before the filing, so here are the dates in order:
- January 2023: Google introduces MusicLM.
- January 2024: Billboard reports Google trained its music model on major-label recordings, the complaint says.
- June 2024: the RIAA sues Udio, which four former core Lyria contributors left Google to launch.
- February 18, 2026: Google launches Lyria 3.
- March 6, 2026: independent musicians sue Google over Lyria 3 in the same Illinois court (MBW).
- June 8, 2026: Google moves to dismiss, citing YouTube’s terms of service (MBW).
- June 25, 2026: Kent Walker, Alphabet’s Chief Legal Officer and a future defendant, publishes a Google policy paper arguing that training on public web data is fair use (MBW).
- August 14, 2026: Rosen files a similar derivative suit against Apple’s leadership (Bloomberg Law).
- October 30, 2026: Rosen files the Alphabet shareholder lawsuit in Chicago.
So Walker is now a defendant in a case about the practice his own paper defended four months earlier. That’s awkward. Not legally decisive, just awkward.
Lyria, MusicLM and 280,000 hours with the labels taken off
The complaint says Google copied “millions of copyrighted sound recordings, musical compositions, and lyrics” to build Lyria, Lyria 2, Lyria 3, Mulan and MusicLM. It also describes how. Google allegedly pulled audio from recordings, turned it into files a machine can read and cut it into clips for training. Then it “stripped out all copyright information that identified the artists.”
MusicLM, the complaint says, learned from five million clips totalling 280,000 hours of music. That’s about 32 years of listening without a break. Rosen says Google “did not identify the specific recordings, rights holders, or licenses” in that pile. And when Google did ask, he says, it asked only “the largest and most sophisticated players in the industry and not individual copyright holders.”
Much of the Alphabet shareholder lawsuit’s music section borrows from the March class action by indie musicians, and some passages repeat its wording. That matters, because Google’s main defence there is a contract, not fair use. When you accepted YouTube’s terms, Google argues, you gave a licence to YouTube and its “Affiliates”, and Google is one of them. If a judge agrees, a case about a whole class of artists becomes a case about one checkbox.
(We’ve all ticked that box. None of us read past the second paragraph.)
Proxy statements and a $169.6 billion buyback anchor the Alphabet shareholder lawsuit
The shareholder side of the Alphabet lawsuit rests on two claims: the board said misleading things, then bought back $169.6 billion of stock at prices those words propped up. Alphabet’s 2024 proxy (the document shareholders get before they vote directors back in) said the board got regular updates on “safe, secure, and trustworthy AI development”. The complaint says those lines vanished from the 2025 and 2026 proxies. It argues the directors cut them because they knew the lines never should have been there.
In 2025, fighting a shareholder proposal on AI data use, the board told investors that Google tracks “data origins and usage (auditable lineage)” for every AI asset. Rosen calls that misleading too. The buyback covered 1.147 billion shares between April 2023 and November 2025. The complaint says every purchase came while the defendants knew about hidden infringement, so “the Company paid artificially inflated prices.”
One small error, though. The complaint ties Anthropic’s $1.5 billion payment to the music publishers’ 2023 lawsuit. MBW points out that the money actually settled a separate class action by book authors in September 2025. For a filing whose whole argument is that someone else didn’t check their sources, that’s a choice.
Why an Alphabet shareholder lawsuit matters to people who make videos
AI copyright risk used to be the legal team’s problem. The Alphabet shareholder lawsuit tries to make it the board’s personal problem, with clawbacks of executive pay as the lever. The complaint calls the suits against Suno, Udio and Anthropic “blazing red flags” the leaders should have acted on. If a court buys that even partly, every AI company’s board has a reason to want a clean, documented training set before launch. Not a settlement fund after it.
For creators the effect is practical. The music in your video, your stream or your ad now has a supply chain, and courts and investors want to see it. “An AI made it” doesn’t answer that question. “Here are the licensed sources it was built from” does. We covered a similar fight over YouTube videos in the Runway case that survived dismissal, and the theme is the same: public doesn’t mean free to take.
What we think: the Alphabet shareholder lawsuit pushes the bill upstream
Our view: derivative suits like this Alphabet shareholder lawsuit rarely end with a CEO writing a personal cheque. They often end with new rules at the top. Here that would mean a board committee, a compliance officer, and a training-data policy someone actually signs. That boring outcome is the one that counts. Once a company like Alphabet has to show where its music training data came from, “we’ll pay damages later” stops working as a plan.
Rosen has now run the same playbook against Apple and Alphabet within three months. Win or lose, the threat alone puts licensing on boardroom agendas, earlier than lawyers usually see it. We make a generator that builds tracks from samples artists upload and get paid for. That is the supply chain this complaint says Google skipped. So we’re not neutral, and you should read us that way.
What to do with your music this week
Since the Alphabet shareholder lawsuit is nowhere near decided, the useful move is checking what you already publish. According to Mubert’s own data, last 28 days, about one in four tracks run through our copyright checker came back with a rights holder’s claim, down slightly from the period before. That’s a lot of claims.
Run your back catalogue through a checker before a claim finds it first. Keep the licence or receipt for every track you use, AI-made or not. Receipts age well. And when a tool says its music is “trained on everything”, ask what “everything” includes. Alphabet’s board is now being asked the same question, with more lawyers in the room.
Cover: illustration based on a photo of Sundar Pichai by Photographer: Lukasz Kobus (European Commission), CC BY 4.0.
Who we are
