Google Voice AI Class Action Claims Journalists’ Recordings Were Used to Train AI Tools

By Steve Levine · Updated July 23, 2026

A proposed class action lawsuit filed in Illinois federal court claims that Google extracted biometric voiceprints from the professional recordings of well-known journalists, podcasters, and audiobook narrators — without their consent — and used them to train the company’s voice AI products. The case, *Marin v. Alphabet, Inc.*, No. 1:26-cv-05436, was filed in the U.S. District Court for the Northern District of Illinois on May 11, 2026, and publicly announced about a week later.

The lead plaintiff is veteran Chicago broadcast journalist Carol Marin, and the suit alleges violations of the Illinois Biometric Information Privacy Act (BIPA), among other state laws. The core claim is straightforward: decades of publicly available broadcast footage, podcast episodes, and audiobook narration were allegedly scraped and processed to build voice models powering products like Gemini Live, NotebookLM Audio Overviews, and YouTube auto-dubbing. The plaintiffs argue that a person’s voiceprint is a biometric identifier under Illinois law — no different, legally, from a fingerprint or a face scan — and that Google never obtained the written consent BIPA requires before collecting it. To be clear at the outset: as of July 2026, this case is in its earliest stages. There is no settlement, no judgment, no certified class, and no claims process. Anyone suggesting a payout is currently available from this lawsuit is ahead of the facts.

What Does the Google Voice AI Class Action Actually Allege?
Who the Plaintiffs Are — and Why Their Professions Matter
This Is One of Nine Coordinated Lawsuits Against Big Tech
What Damages Are on the Table — and What BIPA Math Looks Like
The Legal Hurdles This Case Has to Clear
How This Differs From the Copyright Lawsuits Against AI Companies
What Illinois Residents and Voice Professionals Should Do Now
Frequently Asked Questions

What Does the Google Voice AI Class Action Actually Allege?

According to the complaint, Google extracted biometric voiceprints from thousands of hours of publicly available professional recordings and used them to train the voice AI models behind Gemini Live, NotebookLM Audio Overviews, YouTube auto-dubbing, Google Cloud Text-to-Speech, and Google Assistant. The plaintiffs say this happened without notice, without written consent, and without any disclosure of how long the data would be kept — all requirements under BIPA. The legal theory turns on a distinction worth understanding. The recordings themselves were public: anyone could watch a Carol Marin newscast or stream an investigative podcast.

What the plaintiffs say was *not* public, and not free for the taking, is the biometric data embedded in those recordings — the unique vocal characteristics that identify a specific human being. Under BIPA, publicly displaying your face doesn’t authorize a company to run facial recognition on it; the plaintiffs argue the same logic applies to a voice. A useful comparison is the wave of BIPA facial recognition cases, including the Facebook tag-suggestion litigation that settled for $650 million — those cases also involved data users had voluntarily shared, but courts found the biometric extraction itself required consent. Beyond BIPA, the complaint asserts claims under the Illinois Right of Publicity Act, the Illinois Consumer Fraud and Deceptive Business Practices Act, the Uniform Deceptive Trade Practices Act, and a common-law unjust enrichment theory — the argument that Google profited from the commercial value of the plaintiffs’ voices without paying for it.

Who the Plaintiffs Are — and Why Their Professions Matter

The seven named plaintiffs are all people who make their living with their voices. They include broadcast journalists Carol Marin and Philip Rogers, investigative podcasters Yohance Lacour, Alison Flowers, and Robin Amer, and audiobook narrators Lindsey Dorcus and Victoria Nassif. Lacour and Flowers are Pulitzer Prize winners. That roster is not accidental: professional voice workers have both an unusually large public audio footprint (making them attractive training data) and a direct economic stake in AI tools that can synthesize humanlike speech. The plaintiffs are represented by Loevy + Loevy, a Chicago civil rights firm with a substantial class action practice.

The choice of Illinois as the venue matters enormously, because BIPA is the only state biometric privacy law that gives private individuals a broad right to sue with statutory damages attached. A voice actor in Texas or California whose recordings were used the same way would have far weaker claims — a limitation worth flagging for anyone hoping this case signals a nationwide remedy. The proposed classes here are built around Illinois connections, and people outside that framework may have no claim under these statutes at all. There’s also a warning embedded in the professional identity of the plaintiffs. For narrators and podcasters, voice cloning is not an abstract harm: a synthetic voice trained on their work can compete directly against them for the very jobs that generated the training data. That economic displacement argument runs through the unjust enrichment and right of publicity claims.

This Is One of Nine Coordinated Lawsuits Against Big Tech

The Google case does not stand alone. The same group of seven plaintiffs filed nine coordinated proposed class actions against a roster of major technology companies, including Amazon, Apple, Meta, Microsoft, NVIDIA, ElevenLabs, Adobe, and Samsung. Each complaint follows the same basic template: the company allegedly harvested biometric voiceprints from publicly available recordings to train its voice AI systems, in violation of Illinois law. A concrete example of the parallel filings is *Flowers v. Microsoft*, whose complaint is publicly available through Courthouse News.

Reading it alongside the Google complaint shows how the plaintiffs have tailored the allegations to each defendant’s specific products — Microsoft’s speech synthesis tools in one case, NVIDIA’s voice AI models in another, ElevenLabs’ commercial voice cloning platform in a third. Bloomberg Law’s coverage of the NVIDIA filing framed the coordinated campaign as one of the most significant tests yet of whether BIPA reaches AI training data. The coordinated structure matters strategically. If one defendant wins an early dismissal on a legal theory, the others will cite it; if one court lets the BIPA claims proceed, plaintiffs in the parallel cases gain leverage. The nine cases will likely develop unevenly, and outcomes in one will shape settlement calculus in the rest.

What Damages Are on the Table — and What BIPA Math Looks Like

BIPA provides statutory damages of up to $5,000 per intentional or reckless violation and $1,000 per negligent violation. The complaint also seeks injunctive relief: an order requiring Google to stop collecting voiceprints from Illinois recordings without consent and to destroy biometric data it obtained unlawfully. For companies, the destruction remedy can be as consequential as the money, because it potentially reaches trained models built on the contested data. The per-violation structure is what makes BIPA cases financially serious. In past biometric litigation, defendants have argued that damages should be assessed once per person; plaintiffs have argued each scan or extraction is a separate violation, a reading the Illinois Supreme Court endorsed in *Cothron v.

White Castle*, where the exposure math grew so large that the court itself invited legislative correction (the legislature later amended BIPA to temper per-scan accrual). The tradeoff for plaintiffs is familiar: astronomical theoretical exposure tends to produce settlements at a small fraction of the statutory maximum. The Facebook BIPA settlement paid class members several hundred dollars each — not $5,000 — and that case involved a certified class and years of litigation. Anyone tempted to multiply $5,000 by thousands of hours of recordings should treat that number as a negotiating ceiling, not an expected recovery. No court has yet ruled on how BIPA damages would apply to AI training, and Google will contest every element of liability first.

The Legal Hurdles This Case Has to Clear

The lawsuit faces real obstacles, and it would be a mistake to assume BIPA’s plaintiff-friendly reputation guarantees success. Google’s most likely early arguments include: that a voiceprint used for model training is not a “biometric identifier” as BIPA defines it, because the statute contemplates data used to *identify* individuals rather than to teach a model general speech patterns; that BIPA cannot constitutionally reach conduct occurring on servers outside Illinois; and that some claims may be preempted or time-barred. How the Northern District of Illinois resolves the “identifier” question could determine whether any of the nine coordinated cases survive. The American Bar Association’s litigation section has flagged these voiceprint cases as the leading edge of a new trend in biometric privacy litigation precisely because the statutory text was written in 2008, long before generative AI existed.

Courts will have to decide whether a law aimed at fingerprint time clocks and retina scanners stretches to cover training corpora. Similar interpretive fights have gone both ways: BIPA plaintiffs have won expansive readings on accrual and standing, but defendants have succeeded in cases where the connection between the data and individual identification was attenuated. A practical warning for readers: early-stage class actions like this one routinely take three to five years to resolve, and many never reach a class-wide recovery. Motions to dismiss in the coordinated cases will likely be briefed through late 2026 and into 2027 before anyone knows whether the claims are viable.

How This Differs From the Copyright Lawsuits Against AI Companies

Most litigation over AI training data — the newspaper and author suits over text models, the artist suits over image generators — sounds in copyright. This case deliberately takes a different route.

Copyright claims over voice are weak because copyright protects the recording, not the vocal identity within it, and the recordings here may be owned by broadcasters or publishers rather than the speakers. By framing the harm as biometric privacy and right of publicity, the plaintiffs sidestep ownership questions entirely: Carol Marin may not own the copyright in her old newscasts, but under the plaintiffs’ theory she owns her voiceprint no matter who holds the tape. That framing echoes the Illinois Right of Publicity Act claims that voice actors have pressed elsewhere, and it explains why these nine suits were filed in Chicago rather than the Northern District of California, where most AI copyright cases are pending.

What Illinois Residents and Voice Professionals Should Do Now

Because there is no settlement or claims process, there is nothing to file and no registration deadline — and no legitimate reason for anyone to ask for your personal information in connection with this case. Be skeptical of websites or social media posts promising Google voice AI payouts; claim-aggregation scams routinely appear within weeks of high-profile filings like this one. People who believe their recordings may have been used — particularly Illinois-based journalists, podcasters, narrators, and other voice professionals with substantial public audio — can monitor the docket for *Marin v.

Alphabet, Inc.*, No. 1:26-cv-05436 (N.D. Ill.), or contact plaintiffs’ counsel Loevy + Loevy, which maintains a public page describing its BIPA voice AI litigation. If a class is eventually certified and a settlement or judgment follows, court-approved notice would go out at that point, with instructions and deadlines set by the court.

Frequently Asked Questions

Is there a settlement or payout available in the Google voice AI lawsuit?

No. As of July 2026, the case is in early litigation. There is no settlement, judgment, class certification, or claims process, and any site claiming otherwise is wrong.

What law is the lawsuit based on?

Primarily the Illinois Biometric Information Privacy Act, along with the Illinois Right of Publicity Act, the state Consumer Fraud Act, the Uniform Deceptive Trade Practices Act, and unjust enrichment.

Who filed the lawsuit?

Seven Illinois voice professionals — including journalists Carol Marin and Philip Rogers and Pulitzer winners Yohance Lacour and Alison Flowers — represented by the Chicago firm Loevy + Loevy.

What damages does BIPA allow?

Up to $5,000 per intentional or reckless violation and $1,000 per negligent violation, plus injunctive relief. Actual recoveries in past BIPA settlements have been far lower per person.

Is Google the only company being sued?

No. It is one of nine coordinated proposed class actions by the same plaintiffs against companies including Amazon, Apple, Meta, Microsoft, NVIDIA, ElevenLabs, Adobe, and Samsung.

Do I qualify if I live outside Illinois?

Probably not under these claims. BIPA is an Illinois statute, and the proposed classes are built around Illinois connections.

Browse every open class action settlement at OpenClassActions. Enter free giveaways and sweepstakes at Giveaway Goose. Forgot the name of a movie? Identify it at FindThisMovie. Caring for someone with dementia? Find practical guides at HelpDementia. Watching prices and your paycheck? Follow the numbers at Inflation Money.