Andersen v. Stability AI: The Artists Class Action

In January 2023, three working artists did something that sounded almost quixotic: they sued the companies behind the most popular AI image generators on the planet. Cartoonist Sarah Andersen, painter Kelly McKernan, and concept artist Karla Ortiz filed a class-action lawsuit accusing Stability AI, Midjourney, and DeviantArt of scraping their art — and the art of millions of others — to train image generators like Stable Diffusion, all without a licence, a credit, or a cent.

It was the first case of its kind: not a corporation defending its catalogue, but ordinary creators banding together to ask a simple question. If a machine learned to imitate you by ingesting your life’s work without asking, is that legal? Three years on, that case — Andersen v. Stability AI — is still going, and it may end up being the one that answers the question for everyone. Here’s what’s actually happening, in plain English.

Three small gold artist figures standing together facing a towering AI neural network, symbolising artists taking on AI companies — A handful of working artists versus the biggest names in AI image generation.

The 30-second version

A group of visual artists is suing several AI companies in federal court in California, claiming their copyrighted art was copied to build AI image generators that now compete with them. The case is called a class action because the named artists are trying to represent a much larger group of creators whose work was allegedly used the same way.

Track the cases yourself

Every current AI-copyright ruling — and the ones still being fought — lives in our AI Copyright Ruling Tracker. Filter by your creator type to see, in plain language, how each case affects your rights.

So far, the court has thrown out some of the artists’ claims but kept the most important ones alive — including the core allegation that copying their work to train the models was copyright infringement. The case is now in discovery (the evidence-gathering phase), with a trial scheduled to begin in September 2026. Critically, no judge has yet ruled on the biggest question of all: whether training AI on copyrighted art counts as fair use.

Who is suing, and what they claim

The lead plaintiffs are real, recognisable artists — Andersen is known for the webcomic Sarah’s Scribbles — and more creators joined as the case grew. The defendants are a who’s-who of AI image tools: Stability AI (Stable Diffusion), Midjourney, DeviantArt (which built a generator called DreamUp on top of Stable Diffusion), and later Runway AI.

The artists’ complaint threw several theories at the wall:

Direct copyright infringement — that their images were copied wholesale to assemble the training data.
Induced infringement — that the companies built and distributed tools designed to churn out infringing images.
DMCA violations — that the companies stripped out or altered the “copyright management information” (the credit and ownership data attached to images).
Right of publicity and trademark/trade-dress claims — tied to the tools letting users generate images “in the style of” a named living artist.

The heart of it is straightforward: the artists say their work is the raw material these products were built from, and they were never asked.

A gold timeline with some milestones glowing and one struck through, symbolising claims that survived and were dismissed — Some claims were thrown out early; the core copyright claims survived and are heading toward trial.

What the court has actually decided so far

This is where the headlines tend to mislead, so read carefully. The judge, William Orrick of the Northern District of California, has issued two important early rulings — and neither one is a final verdict on anything.

On 30 October 2023, he largely granted the companies’ motions to dismiss the first version of the complaint — but he let the crucial piece survive: Andersen’s direct copyright infringement claim against Stability AI. He also gave the artists permission to fix and refile the rest.

They did, and on 12 August 2024, Orrick issued a more consequential order. He dismissed the DMCA claims for good (“with prejudice,” meaning they can’t come back). But he allowed the big claims to proceed — direct copyright infringement, induced infringement, and the trademark/trade-dress claims all survived.

One passage from that ruling matters more than the rest. The AI companies argued their tools were like the VCR — a general-purpose technology capable of both legal and illegal uses, and therefore protected. Orrick wasn’t persuaded. A VCR, he noted, doesn’t come pre-loaded with copyrighted works; these models, the artists plausibly allege, were “built to a significant extent on copyrighted works,” and the way they operate “necessarily invokes copies or protected elements of those works.” Whether that’s ultimately true, he said, “depends on how Stable Diffusion works” — a question for later, after discovery. Translation: you don’t get to wave this away by comparing yourself to a tape deck.

The huge question that’s still open

Here’s the part to hold onto: the court has not decided whether training AI on copyrighted art is legal. Surviving a motion to dismiss only means the artists’ claims are strong enough to move forward and be tested with evidence. It is not a ruling that the companies did anything wrong.

The companies’ main defence — the one everything hinges on — is fair use: the argument that training a model is a “transformative” use that doesn’t require permission, similar to how search engines index the web. The artists say it’s nothing of the sort; it’s industrial-scale copying that produces a direct competitor to the very people whose work was taken. That collision hasn’t been resolved here. It’s the fight the trial is built around.

A gold question mark formed from a tilting balance scale weighing a paintbrush against an AI model, symbolising the unresolved fair-use question — The biggest question — is training on your art ‘fair use’? — is still unanswered.

And Andersen isn’t happening in a vacuum. It was the first creator class action of its kind, but there are now well over a dozen similar suits — authors, musicians, and other artists against different AI companies. Some have produced conflicting signals. That’s exactly why a full trial in a case like this one could become a landmark: it would be one of the first times a jury and judge examine, in detail and with real evidence, how these models are built and whether that building infringes.

Why this case matters even if you’re not in it

You didn’t file this lawsuit, and you’re probably not a named plaintiff. So why care? Because the outcome sets the ground rules for everyone who makes things.

If the artists win — or force a meaningful settlement — AI companies may have to license training data, credit creators, or pay for what they use. That reshapes the economics of every generator you’ve heard of. If the companies win on fair use, the message to the industry becomes: your public portfolio is fair game for training, full stop. Either way, the answer will ripple through licensing deals, platform terms of service, and the next decade of tools. This is one of the cases writing the rulebook you’ll have to live under.

What this means for you — and what to do now

You can’t control a federal lawsuit, but you can act on what it teaches:

Don’t trust the headlines. “Judge lets AI lawsuit proceed” is not “artists won,” and “claims dismissed” is not “AI is legal.” The central question is genuinely undecided — be skeptical of anyone who tells you it’s settled.
Keep provenance records. Save originals, working files, timestamps, and posting history. If the law ever gives creators a real claim, the people who can prove a work is theirs — and when they made it — will be the ones who can act on it.
Register the copyrights that matter. In the US, registration is what unlocks the strongest remedies (including statutory damages) if you ever need to sue. Your most valuable pieces are worth the modest filing fee.
Read the terms before you upload. Platforms like DeviantArt were pulled into this precisely over how user art fed into AI features. Know what rights you’re granting when you post.
Watch the fair-use rulings, not the noise. The Andersen trial and its cousins are where creators’ actual rights will be defined. That’s the signal worth following.

A gold shield in front of signed artwork and timestamped files, symbolising protecting your work with provenance — Your best defence is unglamorous: records, timestamps, and proof you made it first.

Andersen v. Stability AI began as three artists refusing to accept that their work could be quietly absorbed into a machine that competes with them. Whether they ultimately win or lose, they turned a private grievance into the case that could decide the rules for every creator in the AI era. That’s worth paying attention to — even if you never set foot in a courtroom.

This article is general information, not legal advice. AI and copyright law is moving fast and differs by country and state — for a specific situation, check the current law where you are or talk to a qualified attorney.

IPSurge helps creators and small businesses understand and protect their intellectual property — the ideas, names, and work that make them unique.

Sources & further reading:

Andersen v. Stability AI: The Artists’ Class Action, Explained