Can a 2008 law written for fingerprint scanners and time clocks reach the billions of voice clips scraped to train generative AI? Nine defendants filed motions to dismiss in Chicago federal court arguing no — and the answer will matter far beyond Illinois. The lawsuits, filed in May 2026 by journalists, podcasters and audiobook narrators against Apple, Amazon, Meta, Microsoft, Nvidia, Samsung, Alphabet, Adobe and ElevenLabs, skip copyright entirely and lean on the Biometric Information Privacy Act’s consent-and-disclosure regime, plus state publicity, consumer-fraud and deceptive-practices claims, according to reporting synthesized from The American Bazaar, jdjournal.com and Мера.
The evidence so far cuts both ways. Plaintiffs’ counsel, Ross Kimbarovsky of Loevy & Loevy, told reporters the firm gathered recordings from podcasts, broadcasts, YouTube and audiobooks and alleges the companies extracted “voiceprints” for AI training — but as jdjournal.com notes, the core legal fight is whether a voiceprint derived from public audio counts as biometric identifier under BIPA at all, and whether the conduct occurred “in Illinois” as the statute requires. Apple’s lawyers countered that mere online availability of a recording proves nothing about whether it was actually ingested into a training set, a causation gap that has sunk BIPA claims before.
The case turns less on whether voice AI used these recordings than on whether a statute built for fingerprint scanners can even see a voiceprint as biometric data in the first place.
The stakes are concrete. BIPA’s $5,000-per-intentional-violation penalty, multiplied across potentially millions of clips and nine defendants, is why Kimbarovsky is pegging exposure at hundreds of millions of dollars per company — a number that echoes Loevy & Loevy’s own track record, including a $75 million BNSF settlement in 2024 and a $47.5 million Motorola deal in 2025, both under the same statute. That history is precisely why the defense bar assembled here — Gibson Dunn, Latham & Watkins, Quinn Emanuel, Sidley Austin and others — is arguing procedure as hard as substance, with Apple pushing to consolidate all nine cases (currently scattered across seven judges) before deciding merits.
For data-sourcing teams, the real signal is structural: this is a privacy-statute attack on training-data pipelines that doesn’t require proving copyright infringement, only that a biometric identifier was extracted without notice. If a judge finds voiceprints qualify as biometric data under BIPA regardless of where the underlying audio was posted, every company running voice-cloning or dubbing products on scraped audio — not just the nine named here — inherits a new compliance surface. Watch for the consolidation ruling and any early decision on the Illinois-nexus argument; either could set the template for whether BIPA becomes AI training’s next copyright-style liability front.
They allege only that their recordings are available on the internet; they do not offer any facts to support an inference that Apple used those recordings to train its technology.
— Reuters
