Kogon v. Google, LLC
Issue
Whether Google's unauthorized reproduction and commercial exploitation of copyrighted sound recordings, musical compositions, and lyrics to train its Lyria AI music-generation systems constitutes direct, contributory, and vicarious copyright infringement under 17 U.S.C. § 501, and whether Google's stripping of copyright management information during its training pipeline violates 17 U.S.C. §§ 1201 and 1202 of the DMCA.
What Happened
Plaintiffs — a putative class of independent musicians and songwriters — filed this complaint in the Northern District of Illinois against Google LLC, alleging that Google copied tens of millions of copyrighted sound recordings, compositions, and lyrics to train its MuLan, MusicLM, and Lyria AI music-generation models without license, consent, or compensation. The complaint alleges that Google's training pipeline stripped identifying copyright management information (including ISRC codes and ownership metadata) from ingested works, and that Google then commercialized the resulting systems — including Lyria 3, launched February 18, 2026, and ProducerAI, launched February 24, 2026 — to over 750 million users. In addition to copyright and DMCA claims, Plaintiffs assert Lanham Act false endorsement and false advertising claims, Illinois BIPA claims for extraction of vocal biometric data without consent, Illinois Right of Publicity Act claims, and Illinois consumer protection and unjust enrichment claims, seeking damages, disgorgement of profits, and injunctive relief.
Why It Matters
This complaint presents a direct test of whether unauthorized ingestion and retention of copyrighted works for iterative AI model training — across successive model generations — constitutes ongoing, compounding infringement rather than a single discrete copying event, a question courts have not yet resolved at scale in the music context. The case is also notable for combining copyright and DMCA claims with biometric privacy and right-of-publicity theories premised on vocal identity extraction, potentially establishing a multi-theory liability framework for AI developers that operates independently of any Section 230 defense.
How accurate was this summary?