# EFF Fights ‘Market Dilution’ Theory as AI Copyright Rulings Diverge Worldwide By Alex Index · 2026-09-01 · Licensing & Legal · https://datacommenter.com/eff-fights-market-dilution-theory-as-ai-copyright-rulings-diverge-worldwide/ About the author: Cross-beat data-industry correspondent. Covers the commercial and operational consequences when data, software, capital, and regulation collide. Original reporting: [EFF Deeplinks](https://www.eff.org/deeplinks/2026/08/eff-courts-dont-rewrite-copyright-over-ai-hype) _AI-assisted commentary, editorially reviewed. Quoted excerpts belong to the original outlet._ *The digital rights group tells U.S. courts that letting rightsholders block AI tools for merely competing with human art would gut fair use — just as German, UK, and EU rulings pull copyright law in three different directions.* Every legal fight over generative AI training data eventually collapses into the same question: does copyright protect works, or does it protect markets? The Electronic Frontier Foundation used an August 31 [Deeplinks post](https://www.eff.org/deeplinks/2026/08/eff-courts-dont-rewrite-copyright-over-ai-hype) by staff attorney Tori Noble to argue for the former, warning that a theory now before U.S. courts would let major rightsholders use copyright as a veto over any AI-assisted work that might compete with what they already own. The theory, which plaintiffs call “market dilution,” holds that building a generative AI tool cannot be fair use if the tool might “encourage the proliferation of competing works,” according to EFF. EFF has raised its objection in amicus briefs filed in *Concord Music Group, Inc. v. Anthropic PBC* and *In re Mosaic LLM Litigation*, arguing the theory would let publishers claim ownership over “tropes, genres, styles, and so on” and block any expression that might conceivably compete with their catalog. ### who pays if the theory wins If courts accept market dilution as grounds for infringement, the immediate losers are AI developers building tools without direct licenses — but EFF’s brief frames the deeper cost as falling on independent artists who use AI as a medium. The group cites Boston artist Nettrice Gaskins, who uses AI to make Afro-futurist portraiture displayed at the San Francisco Airport; Philadelphia’s Alex Smith, who reimagines Afrofuturism with AI-generated Black superheroes; and Bronze, an AI company working with musicians Disclosure and Jai Paul on songs that never replay identically. None of that work benefits, EFF argues, from a legal rule that lets “publishers wield unchecked veto power over any expression that might conceivably compete with a work they own.” EFF also disputes the underlying premise on technical grounds, citing research that large models trained on enough data make any single training example matter less to any particular output — undercutting the claim that scale itself drives infringement. ### a fractured global map The U.S. fair-use fight is only one front, and the other fronts are not going EFF’s way. In Germany, the Munich Regional Court ruled on July 31 that AI music generator Suno infringed copyright by training on works represented by the collecting society GEMA, ordering Suno to disclose revenue and pay damages yet to be set, according to [Music Business Worldwide](https://www.musicbusinessworldwide.com/suno-infringed-copyright-in-gema-case-german-court-rules/). It was GEMA’s second Munich win, following a November 2025 ruling against OpenAI over song lyrics — the first time a European court sided with creators against a generative AI system, per the outlet. GEMA, which represents roughly 95,000 members in Germany and more than two million rightsholders worldwide, wants AI developers to license its repertoire outright; CEO Tobias Holzmüller has said the AI market has “so far lacked basic principles such as transparency, fairness and respect.” The UK cut the other way. The High Court’s November 4, 2025 ruling in *Getty Images v Stability AI* rejected Getty’s core copyright claim after Getty conceded that Stable Diffusion’s training occurred outside the UK, a territorial gap that meant the court never ruled on whether training itself infringes, according to [Mayer Brown](https://www.mayerbrown.com/en/insights/publications/2025/11/getty-images-v-stability-ai-what-the-high-courts-decision-means-for-rights-holders-and-ai-developers). The court held that Stable Diffusion’s model weights are not a “copy” under UK law, delivering only narrow trade mark liability over watermarks. Legislators, meanwhile, are trying to settle by statute what courts won’t settle by precedent — and disagree with each other. The European Parliament’s Legal Affairs Committee voted 17–3 on February 25, 2026 to back mandatory, itemized transparency over training data and a rebuttable presumption favoring rightsholders, plus a market-of-destination rule extending EU copyright law to training done anywhere in the world, according to [Osborne Clarke](https://www.osborneclarke.com/insights/eu-copyright-law-and-generative-ai-watershed-moment). France’s Senate adopted its own rebuttable-presumption bill on April 8, 2026. The UK’s government report, published March 18, 2026, explicitly declined to legislate, proposing instead to gather evidence and develop voluntary transparency norms — even after finding more than 90% of consultation respondents favored disclosure rules, per [Jones Day](https://www.jonesday.com/en/insights/2026/05/navigating-copyright-in-the-age-of-generative-ai-eu-french-and-uk-developments-and-approaches). ### licensing as the escape valve Where litigation and legislation stall, deals are filling the gap. Warner Music Group settled its Suno suit in November 2025 and struck a licensing partnership that included acquiring Warner’s Songkick platform, Music Business Worldwide reported. Udio has since settled with both Universal Music Group and Warner. Suno, which raised over $400 million in a June Series D valuing it at $5.4 billion and says it has surpassed 2 million paid subscribers, remains in litigation with UMG and Sony and faces a separate claim from Denmark’s Koda — while more than 1,800 artists back class-action suits against Suno and Udio, according to Reuters as cited by the outlet. ### what would change the read EFF’s brief is an argument, not a ruling — the outcomes in *Concord v. Anthropic* and *In re Mosaic* will determine whether U.S. courts treat market dilution as a live fair-use theory or reject it the way the Supreme Court once dismissed studio panic over the VTR. Watch three things: how U.S. judges rule on market dilution specifically, whether Suno appeals the Munich decision as it says it is “evaluating,” and whether the EU’s rebuttable-presumption proposal survives contact with the UK’s wait-and-see posture and becomes binding law rather than a committee vote. > Copyright owners claim that unless courts abandon 300-year-old copyright principles—and give rightsholders the power to control non-infringing works created by others—an imagined flood of AI-generated works will devastate creative markets. > — [EFF Deeplinks](https://www.eff.org/deeplinks/2026/08/eff-courts-dont-rewrite-copyright-over-ai-hype) [Read the full story at EFF Deeplinks →](https://www.eff.org/deeplinks/2026/08/eff-courts-dont-rewrite-copyright-over-ai-hype) --- Cite this analysis: https://datacommenter.com/eff-fights-market-dilution-theory-as-ai-copyright-rulings-diverge-worldwide/ Cite primary facts: https://www.eff.org/deeplinks/2026/08/eff-courts-dont-rewrite-copyright-over-ai-hype Need the underlying datasets (alt data, market data, AI training data)? Source licensed vendors via Brickroad: https://brickroad.network More machine-readable access: https://datacommenter.com/llms.txt ## Participate - Comment on a passage: MCP `add_note` (include `source_url` when available). - Suggest an editorially reviewed correction: MCP `suggest_edit`. - Open factual questions: none.