Stock Photography Giants Are Suing AI Image Generators — And Midjourney Is in the Crosshairs
Copyright lawsuits targeting AI image generators are escalating, and Midjourney sits squarely in the debate over fair use and training data rights.
The legal reckoning that photographers and stock agencies have been demanding for years is no longer theoretical. Getty Images fired the first major shot in February 2023, suing Stability AI in U.S. District Court over the alleged unauthorized use of its library to train AI image models. That case opened a door — and a long line of claimants has been walking through it ever since. Midjourney, one of the most popular AI image generators on the market, hasn’t escaped this scrutiny.
The core dispute is simple, even if the legal resolution isn’t: stock photo companies built their entire business on licensing images, and AI companies trained their models on those same images without paying a cent. Whether that constitutes copyright infringement or falls under fair use is a question courts are still very much working out.
Getty Images Made the First Move
When Getty Images filed against Stability AI on February 3, 2023, the company’s general counsel Craig Peters didn’t mince words. “Stability AI chose to copy our entire portfolio of 75 million images without permission, without payment and without credit,” Peters said in the official company statement. The complaint alleged that Stability AI scraped billions of copyrighted images from Getty’s platform to train its Stable Diffusion model — Getty’s watermark famously appearing in some of the model’s outputs, which did not help Stability AI’s position.
That lawsuit became a landmark moment. It established that major rights holders were willing to fight this in court rather than simply negotiate licensing deals after the fact. Shutterstock took a different approach, striking licensing agreements with AI companies including OpenAI and exploring partnerships rather than litigation — a contrast in strategy that reflects just how unsettled the industry’s response remains.
Why Midjourney Is a Central Name in This Debate
Midjourney has operated in something of a legal gray zone since launch. The platform trained on massive image datasets scraped from the web, a practice shared by virtually every AI image generator. Unlike Stability AI, Midjourney is a private company with no external investors until its $160 million Series C in June 2023 — a funding round that gave it both resources and a higher profile to defend. Artists and photographers have singled out Midjourney repeatedly in public complaints, noting that its outputs often reproduce styles and visual elements tied to specific photographers’ work.
A group of artists, including Kelly McKernan and Karla Ortiz, filed a class-action lawsuit against Midjourney, Stability AI, and DeviantArt in early 2023. The case raised direct questions about whether generating images in an artist’s “style” — even without reproducing a specific work — constitutes infringement. A federal judge dismissed several claims but allowed others to proceed, meaning the legal exposure isn’t going away.
The Fair Use Question Nobody Can Answer Yet
Midjourney’s likely defense, if pushed to court, is fair use — the same argument every AI company in this space is expected to make. The argument runs roughly as follows: training a model on images is transformative because the model doesn’t store the images, it learns statistical patterns from them. The output is a new creation, not a reproduction.
Legal scholars aren’t convinced this is settled law. Fair use analysis in the U.S. involves four factors, and courts weigh them case by case. The transformative use argument has worked in some contexts — Google Books scanning millions of books, for example — but AI training data involves commercial products generating revenue in direct competition with the original creators. That’s a harder case to make, and no court has definitively ruled on AI training data as fair use.
The broader creative industry is watching this closely. If a major ruling comes down against AI companies, the entire model-training pipeline for image generators would need to change — requiring licensed datasets, revenue sharing agreements, or opt-in consent from rights holders.
What This Means for Midjourney Users
For now, Midjourney keeps shipping updates and its user base keeps growing. V7 is the current model, producing outputs that have made earlier versions feel like rough drafts. But the legal cloud overhead isn’t decorative. If courts ultimately require AI companies to pay licensing fees retroactively or going forward, those costs land somewhere — and “somewhere” usually means users, pricing, or both.
The more immediate concern for photographers and stock agencies is precedent. A single ruling clarifying that AI training on copyrighted images without consent is infringement would transform licensing negotiations across the entire industry overnight. Shutterstock’s bet-hedging strategy — licensing deals now rather than lawsuits later — starts to look pretty smart if that ruling comes through. Getty Images, meanwhile, is still in court and playing for keeps. Midjourney isn’t the named defendant in that case, but it’s watching every word of the judgment.
