I keep this map open in a tab.

Not because I need to track every case. I don't. But because there's something clarifying about seeing it all at once, the sheer volume of litigation that has accumulated since 2022, plotted across the country like a weather system. The version embedded below is dated March 5, 2026. By the site's April 3, 2026 update, the count had reached roughly 100 US suits, with about 30 more filed abroad. The number will be higher by the time you read this.

The map was made by Professor Edward Lee's ChatGPTiseatingtheworld.com and shared under a Creative Commons license. I'm embedding it here with attribution because every independent designer working with AI tools, or whose work has been scraped without permission, should know what the legal landscape actually looks like right now.

What the map shows

The heaviest dockets are in the Northern District of California and the Southern District of New York. That's not surprising. Those are where the tech companies are headquartered and where the publishers and entertainment companies have their lawyers. But the spread matters. Cases have been filed in Delaware, Colorado, Tennessee, and beyond, and several have transferred between districts. The first appeal on the core question has already reached a circuit court: the Third Circuit heard argument on June 11, 2026 in Thomson Reuters v. Ross Intelligence, reviewing the February 2025 ruling that rejected fair use for an AI legal research tool trained on Westlaw headnotes. That means binding precedent is coming.

A few cases to know by name if you're in fashion and adjacent creative industries:

Andersen v. Stability AI (N.D. Cal., Judge Orrick). This is the visual artists' case. Sarah Andersen and a class of illustrators sued Stability, Midjourney, and others over image generation models trained on their work without consent or compensation. The court trimmed the case in 2023, but in August 2024 it let the core copyright claims, and trade dress claims against Midjourney, move into discovery. Trial is currently set for September 2026. What survives will tell image creators what copyright law will and won't protect.

Getty Images v. Stability AI (US and UK). Getty went after Stability directly, in part over the reproduction of its watermarks in generated images. The UK went first: after Getty dropped its core training claims mid-trial for jurisdictional reasons, the High Court's November 4, 2025 judgment rejected the secondary copyright claim and gave Getty only very limited trademark findings. The US case was voluntarily dismissed in Delaware and refiled in the Northern District of California in August 2025, where it survived a motion to dismiss in part this April. If your brand has a signature visual identity, a logo, a pattern, a recognizable aesthetic, this pair of cases is the closest legal analogue to what happens when that gets absorbed into a generative model.

Bartz v. Anthropic (N.D. Cal.). The authors' case against the company that makes the AI I'm using right now, and the source of the biggest number in this entire map: a $1.5 billion class settlement over books downloaded from pirate libraries. Judge Alsup ruled in June 2025 that training on lawfully acquired books was fair use but that building a library of pirated copies was not. The settlement got preliminary approval in September 2025; final approval was still pending after the May 14, 2026 fairness hearing. Separately, music publishers including UMG and Concord are litigating against Anthropic over song lyrics, with new suits filed in early 2026 covering more than 20,000 compositions.

Kadrey v. Meta (N.D. Cal., Judge Chhabria). The mirror image of Bartz. In June 2025 the court granted Meta summary judgment on fair use for training, but pointedly said the ruling reflected these plaintiffs' failure to prove market harm, not a green light for the industry. The fight over Meta's alleged uploading of pirated books continues.

In re OpenAI Copyright Infringement Litigation (S.D.N.Y., MDL No. 3143, Judge Stein). In April 2025 the Judicial Panel on Multidistrict Litigation consolidated the OpenAI and Microsoft cases, including the writers' suits from Tremblay, Silverman, and Chabon, the Authors Guild case, and The New York Times case, into a single proceeding in Manhattan. Motions to dismiss the output-based infringement claims have largely failed, and discovery is enormous: in January 2026 the court affirmed an order requiring OpenAI to hand over 20 million de-identified ChatGPT logs.

Disney v. Midjourney (C.D. Cal.). Disney and Universal sued in June 2025 over image outputs of their characters, and Warner Bros. followed. This is the case to watch on outputs rather than training, because the studios' theory, that a generator which produces recognizable protected characters infringes, is structurally the theory a designer would use if a model started producing recognizable versions of a signature print or silhouette.

What it means for you

The honest answer is: the law is unsettled, but it is no longer silent. The first rulings on whether training is fair use have arrived, and they conflict. One judge said training on lawfully bought books is fair use. Another said fair use, but only because the plaintiffs couldn't prove market harm. A third, in the Ross case, rejected fair use outright. The courts still haven't settled what "substantial similarity" means when the output is probabilistic rather than directly copied, and they haven't decided whether a style can be protected, which is the question that matters most for fashion designers.

What we do know:

The volume of litigation means the law is going to move fast in the next two to three years. The first circuit-level decisions are coming. If you're building a brand identity, a print, a silhouette language, a signature technique, now is the time to document it. Dates, process records, publication evidence. Copyright registration is cheap and the evidentiary value is significant if you ever need to prove priority. My provenance tool exists for exactly this.

The California cases are being watched closely for another reason: California's AI statutes, AB 2602 and AB 1836 on digital replicas, and the AB 2013 training data transparency law now in effect, create a parallel statutory track that may give creators remedies even where copyright claims fail.

And if you've licensed your work to stock platforms, or posted high-resolution images of your collections publicly, your designs are almost certainly in at least one of these companies' training sets. That's not a legal conclusion, it's a probability statement, and I explain the mechanics in Your Designs Are Probably in an AI Training Dataset. The litigation will eventually tell us whether that fact entitles you to anything.

Use the map

Zoom in. The interactive embed below lets you explore by district. Click through to the original at ChatGPTiseatingtheworld.com for case links and updates, they maintain it actively. It's the best single resource tracking this litigation in real time, and I keep the fashion-relevant developments in the tracker.

The map is licensed under Creative Commons (CC BY 4.0). Attribution: ChatGPTiseatingtheworld.com, March 5, 2026.

Sources

This article is editorial analysis, not legal advice. For questions about your specific situation, consult a qualified attorney.