In "EFF to Courts: Don't Rewrite Copyright over AI Hype," published August 31, 2026, the Electronic Frontier Foundation urges federal judges not to rewrite three centuries of copyright doctrine because AI has people spooked. The post, written by EFF attorneys Tori Noble and Corynne McSherry, lays out arguments from the group's amicus briefs in Concord Music Group, Inc. v. Anthropic PBC and In re Mosaic LLM Litigation.
Short answer: EFF is asking courts not to expand copyright protection because of AI hype. In amicus briefs filed in Concord Music Group v. Anthropic and In re Mosaic LLM Litigation, EFF argues that publishers' "market dilution" theory — treating AI-assisted competition itself as copyright harm — would gut fair use, not just for AI companies but for everyone who builds on existing work.
In my testing of the primary documents behind this story, I read the EFF post directly rather than the aggregator writeups quoting it secondhand, then cross-checked the two lawsuits it names against their public court dockets. The underlying claims held up. This piece walks through what EFF is actually arguing, what the two cases are about, and how to read a legal filing like this one without assuming it settles anything — because neither case has gone to trial yet.
What you'll need to understand this fight
You don't need a law degree to follow this, but a few terms matter. "Fair use" is the legal doctrine that lets you use copyrighted material without permission under certain conditions — commentary, parody, and yes, some machine learning training have all been argued under it. An "amicus brief" is a filing from someone who isn't a party to a lawsuit but wants to give the court their view; EFF isn't suing anyone here, it's weighing in. And "market dilution" is the theory at the center of this dispute: publishers argue that AI tools deserve less fair-use protection because they might help create more works that compete with the originals. EFF's position, boiled down, is that competition has never been a copyright harm on its own — otherwise a human author who reads a hundred novels and writes a better one would be infringing too.
Step-by-step: how to read EFF's argument for yourself
1. Start with the primary source, not a summary of a summary
Go straight to EFF’s own post instead of a social-media recap. Recaps tend to compress "EFF opposes an expansion of copyright doctrine in two specific cases" into "EFF says AI training is legal," which isn't what the brief claims. EFF is defending existing fair-use limits, not making a blanket statement about every AI use case.
2. Identify which cases the brief actually addresses
EFF's arguments apply to two live disputes: a music-publisher suit against an AI company, and a group of authors suing over training data. Neither is decided. Anything you read that treats this as a finished ruling is getting ahead of the record.
3. Pin down the "market dilution" argument specifically
The publishers' theory in these cases isn't the usual fair-use fight over whether a specific output copies a specific work. It's broader: that AI tools shouldn't get fair-use protection because they might enable more competing creative works generally. EFF's brief calls this a break from how copyright has always worked — copyright protects specific expression, not a rightsholder's market position against new competitors.
4. Check the historical precedent EFF leans on
EFF points back to Sony Corp. of America v. Universal City Studios, Inc., the 1984 "Betamax" case, where the Supreme Court declined to treat the VCR as inherently infringing because it had substantial legitimate uses like recording a show to watch later. Industry lobbyists at the time predicted the VCR would gut the film business — MPAA president Jack Valenti told Congress in 1982 that "the VCR is to the American film producer and the American public as the Boston strangler is to the woman home alone." The format ended up creating a home-video rental market worth billions to those same studios.
5. Watch what the courts actually decide, not what either side predicts
Both cases are still in litigation. EFF's brief is an argument for judges to weigh, not a verdict. If you're tracking the outcome, the docket — not a blog post, including this one — is the record that matters.
The two cases, side by side
| Concord Music Group, Inc. v. Anthropic PBC | In re Mosaic LLM Litigation | |
|---|---|---|
| Filed | October 18, 2023 (M.D. Tenn., later moved to N.D. Cal.) | N.D. Cal., docket 3:24-cv-01451 |
| Plaintiffs | Universal Music Group, Concord Music Group, ABKCO Music | Authors including Rebecca Makkai and Jason Reynolds |
| Defendant | Anthropic PBC | Databricks, Inc. and its Mosaic ML unit |
| What's alleged | Song lyrics used without permission to train Claude | Books from a Books3-derived dataset used to train the MPT-7B model |
| EFF's role | Filed an amicus brief opposing the market-dilution theory | Filed an amicus brief opposing the market-dilution theory |
| Status as of this writing | Ongoing, no trial verdict | Ongoing, no trial verdict |
Example prompts you can copy
If you want to read a legal filing like this one yourself instead of trusting any summary — mine included — these prompts work well in Claude, ChatGPT, or a research tool once you've pasted in the source text:
- "Read this legal blog post and separate the factual claims (case names, dates, filings) from the opinion or argument being made."
- "What is the 'market dilution' theory this brief is opposing, and how is it different from a standard fair-use defense?"
- "What historical case does this brief compare the current dispute to, and what happened in that earlier case?"
- "List the two lawsuits named here and what each one is specifically about."
- "Draft a two-paragraph plain-English summary of this brief for someone who has never studied copyright law."
I ran the third prompt against the EFF post in both Claude and a NotebookLM notebook built from the source PDF and the case dockets, and got matching answers on the Betamax comparison both times — a decent sanity check when you don't want to take a single tool's summary at face value. My NotebookLM walkthrough and Perplexity research guide cover how I set those up if you want to do the same with your own source documents.
Common mistakes to avoid
The mistake I see most often is assuming EFF represents the AI companies in these cases. It doesn't — EFF is a nonprofit digital-rights group, not a party to either lawsuit, and it has criticized AI companies plenty of times elsewhere. Second, people read "opposes market dilution theory" as "supports unrestricted AI training," which the brief doesn't say; EFF is arguing about one specific legal theory, not blessing every training practice. Third, treating an amicus brief as a ruling — a brief is one filing among many that a judge will weigh, not a decision. Fourth, confusing this dispute with output-plagiarism cases, where the claim is that a model's output copies a specific work; these two cases are about training-data use, a different legal question. And fifth, assuming "fair use" means "no rules apply" — it's a specific four-factor legal test, and courts still have to work through each factor case by case. I've seen the same kind of legal-nuance flattening happen with AI copyright takedowns before, like when a game engine got pulled from Google Play over a baseless AI copyright notice that nobody checked closely before acting on it.
Tools that make this easier
If you're trying to keep up with AI copyright litigation without becoming a legal researcher, a few things help. Feeding primary documents — the brief, the docket entries, the actual complaint — into a tool built for long documents beats skimming secondhand takes; I've tested that workflow in my NotebookLM guide and my Perplexity for research walkthrough, and both hold up better than a general chat prompt for this kind of source-checking. If you want a sense of how AI and copyright law collide outside the US, my piece on why copyright doesn’t protect AI-generated content in the EU covers the flip side of this same debate. And if a specific AI tool has gotten you into a copyright or takedown mess, my rundown of the baseless AI copyright notice that got a project pulled from Google Play and my broader piece on AI mania eviscerating decision-making both cover how often a confident-sounding claim turns out to be thinner than it looks once you check the source.
My take
EFF's brief isn't a prediction that AI companies will win these cases, and it isn't proof that AI training is legal. It's a specific, narrower argument: that "AI might help create more competing work" shouldn't be treated as a new kind of copyright harm, because that logic would let any rightsholder block anything that competes with them, AI or not. Whether the judges in Concord Music Group v. Anthropic and In re Mosaic LLM Litigation agree is still an open question. What's not open to much debate is the historical pattern EFF points to — the same "this technology will destroy the industry" argument got made about VCRs, player pianos, and cameras, and none of those predictions held up the way the industry expected at the time.
Even outside the courtroom, that pattern of confident, unverified predictions is worth watching for. I've flagged the same instinct — treating a prediction as if it were already a fact — in places as different as an AI legal-advice mistake a labor tribunal had to correct and a prompt-injection attempt buried inside an actual legal filing. The common thread in all of these: read the primary source before you repeat the headline.
Frequently Asked Questions
Is EFF a party to the lawsuits against Anthropic and Databricks?
No. EFF filed amicus ("friend of the court") briefs in both cases to argue against a specific legal theory. It isn't suing anyone or being sued, and it isn't paid by either side.
What is the "market dilution" theory EFF is arguing against?
It's the idea that AI tools deserve reduced fair-use protection because they might help produce more works that compete with existing ones. EFF argues this treats ordinary competition as copyright harm, which isn't how the doctrine has ever worked.
Does the 1984 Betamax case actually apply to AI lawsuits?
EFF thinks so, as a precedent for judicial caution: the Supreme Court in Sony Corp. v. Universal City Studios declined to expand copyright liability based on speculation about a new technology's harms, and required proof of substantial non-infringing use instead. Whether today's courts apply the same reasoning to AI training is still undecided.
Have Concord Music Group v. Anthropic or In re Mosaic LLM Litigation been decided?
No. Both cases are ongoing as of this writing, with no trial verdict in either one. EFF's briefs are arguments for the courts to consider, not outcomes.
Where can I read EFF's brief myself instead of a summary?
Start with EFF’s own August 31, 2026 post, which links out to the underlying filings. Reading the primary source first is the single habit that catches most secondhand misreadings.