EFF Fights 'Market Dilution' Theory That Would End Fair Use for AI Training
What happened
The Electronic Frontier Foundation filed amicus briefs in two concurrent U.S. copyright cases -- EFF to Courts: Don't Rewrite Copyright Over AI Hype -- targeting the same novel legal theory in each. In Concord Music Group v. Anthropic, music publishers are pressing claims against the AI lab over training data. In re Mosaic LLM Litigation consolidates related claims in a class format. Both cases feature the 'market dilution' argument, which holds that AI-generated outputs compete with the original works used in training, and that this competition alone satisfies the market harm element of a fair use analysis. The EFF argues that accepting this theory would hand rightsholders a veto over any expressive work that could compete in their market, collapsing the fourth fair use factor into a near-absolute bar. This litigation connects directly to the broader wave of rights-holder suits, including Sony and Warner's claims against Anthropic, that have put training data provenance at the center of AI vendor governance.
Why it matters
- ·If courts accept the market dilution theory, fair use would cease to function as a reliable compliance basis for training data ingestion programs, requiring enterprises and their AI vendors to obtain explicit licenses for a far wider category of content or face direct infringement liability.
- ·Compliance teams that procure foundation models from vendors should reassess whether those vendors have documented their training data legal basis: a court ruling against Anthropic in Concord Music Group could create downstream liability exposure for enterprise deployers who did not conduct adequate due diligence at the intake stage.
- ·Two simultaneous cases proceeding under the same legal theory creates a compounding risk: courts in each case could reach different conclusions, producing conflicting precedents and regulatory uncertainty that cannot be resolved quickly, leaving compliance programs in a holding pattern for months or years.
Governance controls affected
What to do now
- ☐Map every foundation model in your AI inventory to its training data legal basis and flag any that rely solely on undocumented fair use assumptions rather than explicit licensing or formal legal analysis.
- ☐Issue a vendor questionnaire to AI suppliers asking them to specify the legal basis for training data ingestion and whether that basis has been reviewed by counsel in light of the Concord Music Group and Mosaic LLM litigation.
- ☐Engage legal counsel to assess whether your organization's own AI training or fine-tuning activities would survive scrutiny under the market dilution theory, particularly for media, music, publishing, or software content.
- ☐Add Concord Music Group v. Anthropic and In re Mosaic LLM Litigation to your regulatory monitoring tracker and set alerts for material court orders, including rulings on motions to dismiss, summary judgment, and any interlocutory appeals.
- ☐Review vendor contracts to determine whether AI suppliers carry intellectual property indemnification provisions and whether those provisions would cover a ruling that invalidates a fair use defense for training data.
What to watch next
The most consequential near-term signal will be how each court rules on summary judgment motions addressing the fair use question, particularly whether any judge formally endorses or rejects the market dilution framing. A ruling in either case that credits the theory even partially would accelerate licensing demands from rightsholders across the content industry and trigger a revaluation of training data risk in enterprise AI procurement. Compliance teams should also watch for the EU's parallel developments: the EU AI Act and related GPAI training data transparency requirements create a separate but reinforcing pressure on training data documentation that could interact with U.S. litigation outcomes in cross-border AI deployments.
Stay ahead of stories like this
Get every US AI governance development like this one, plus the rest of the week's developments. Every Thursday.