Getty Images' stock went up 150–200% overnight on a deal with OpenAI. That's the headline. But the headline misses the actual story.
Three years ago, Getty was one of the loudest voices demanding that AI companies be held accountable for scraping licensed images without permission. They filed lawsuits in two countries. They became the symbol of "rightsholders fighting back." And now they've signed a multi-year display deal with the largest AI company in the world, agreeing to let their images appear inside ChatGPT.
That flip is the story. Not the stock price.
To understand what it means, you need to know how we got here — and what every other major lawsuit in AI image history actually achieved.
The Lawsuits: A Complete History
November 2022 — Artists vs. Stability AI, Midjourney, DeviantArt
The first major AI image lawsuit came from three artists: Sarah Andersen, Kelly McKernan, and Karla Ortiz. They filed a class-action in the Northern District of California against Stability AI (maker of Stable Diffusion), Midjourney, and DeviantArt's DreamUp.
The claim: billions of images scraped from the internet — including their work — without consent, credit, or compensation, used to train models that then competed with the very artists whose work trained them.
What happened: The case went through multiple rounds of dismissal and refiling. The court dismissed most claims in 2023, finding that the plaintiffs hadn't sufficiently shown that outputs were substantially similar to their specific works. The artists amended and refiled. By 2026, the case is still active against Stability AI and Midjourney in modified form. DeviantArt was largely dropped from the suit after arguing that DreamUp was built on top of Stability AI's model, not an independent scraper.
The lasting impact: This case established the central legal argument that would be used in every subsequent AI image lawsuit — the question of whether training on copyrighted data constitutes infringement, and whether derivative outputs need to be "substantially similar" or whether the training act itself is the violation.
January 2023 — Getty Images vs. Stability AI (UK + US)
This was the case with teeth. Getty filed in the UK first, then in the US District of Delaware, against Stability AI for copyright infringement, trademark dilution, and passing off.
The UK lawsuit was filed under clearer IP law. But the US case had the most explosive evidence: Stable Diffusion was generating images with corrupted, half-visible Getty watermarks embedded in the outputs. There was no ambiguous "was this in the training data?" question — you could literally see the watermark.
Getty's CEO Craig Peters was blunt about the company's position: "We believe artificial intelligence is exciting and we want it to thrive, but it needs to develop responsibly with respect for the legal rights of creators."
What happened: The UK case proceeded. The US case proceeded. Stability AI changed management in 2023 and went through significant turbulence. The litigation continued slowly through 2024 and into 2025. As of the Getty-OpenAI deal announcement in June 2026, Getty's lawsuit against Stability AI is still active — it was never settled, never concluded.
The key detail most people miss: Getty is now licensing its images to OpenAI while still suing Stability AI. These are two different companies and two different legal theories. But the optics are striking — the same Getty that said AI companies needed to respect creator rights is now inside one of the biggest AI products in the world.
2023 — Shutterstock's Different Play
Shutterstock took a different approach entirely. In 2023, they quietly signed a licensing deal with OpenAI to include Shutterstock images in DALL-E training data. At the same time, they launched a "Contributor Fund" to pay photographers whose work was licensed to AI companies.
The AI community's reaction was mixed. Some called it pragmatic. Others called it selling out the artists whose work built Shutterstock's library in the first place. The Contributor Fund's per-image payouts were reportedly negligible.
The lesson Shutterstock demonstrated: You don't have to choose between litigation and surrender. You can license proactively, take the revenue, frame it as "responsible AI development," and point to the Contributor Fund as proof you care about creators. Whether the math works out for individual photographers is a different question.
2023–2024 — The Cascading Cases
The artist lawsuits multiplied:
- Kelly McKernan won a partial procedural victory when the court allowed amended claims forward against Midjourney and Stability AI.
- Concept art studios filed additional suits arguing their house styles were being systematically reproduced ("in the style of [studio name]").
- The Authors Guild vs. OpenAI (text, not images, but same theory) showed how the legal framework was generalizing.
- Stock photography companies beyond Getty started watching the Getty vs. Stability AI litigation closely — it was effectively industry-funded precedent-setting.
Midjourney, notably, made no deals. No licensing agreements, no settlements, no public acknowledgment that the lawsuits were changing their behavior. They kept shipping product. Midjourney V6 launched. Midjourney hardware was announced in June 2026. The litigation continued in the background.
What the Getty-OpenAI Deal Actually Signals
1. Litigation is slow; licensing is fast
Getty's UK case against Stability AI has been running for over three years without resolution. The US case is similarly stalled. Courts move slowly on novel IP questions, and AI companies have strong incentive to delay — every month of delay is another month of building moat, users, and revenue.
OpenAI didn't wait for a lawsuit. They proactively approached Getty (or vice versa) and negotiated terms. The deal closes the litigation risk for OpenAI entirely with respect to Getty's library. Deals are faster and more certain than verdicts.
2. This is a display deal, not a training deal
Read the language carefully: Getty's images will appear "in visual responses inside ChatGPT" for search and discovery. This is a display licensing deal — Getty images showing up when you ask ChatGPT to find you a photo, similar to how Google Images links to licensed sources.
This is meaningfully different from a training data deal. Shutterstock licensed images for training DALL-E. The Getty-OpenAI deal, as announced, is about showing Getty's images in ChatGPT responses — essentially making ChatGPT a discovery surface for Getty's licensed library.
That's a crucial distinction: Getty gets a new distribution channel, OpenAI gets high-quality, licensed, legally-clean visual results for image search queries. No one is saying these images trained anything new.
3. Getty's valuation had collapsed — this is a lifeline
A stock trading at $0.61 is a company the market has given up on. Traditional stock photo agencies have been in freefall since generative AI made it possible to create custom images on demand rather than licensing pre-existing ones. Getty's core business model — you need licensed images, we have them, pay us — was being disrupted by the same AI companies it was suing.
The OpenAI deal doesn't just add revenue. It reframes what Getty is. Instead of a declining stock photo agency being disrupted by AI, Getty becomes a licensed content layer that AI companies need to operate responsibly. That's a completely different valuation story — hence the 200% move.
4. The Midjourney contrast is stark
While Getty is signing deals and Shutterstock has had deals since 2023, Midjourney has made no licensing agreements and reached no settlements. This is either principled defiance or strategic delay, depending on your interpretation.
Midjourney's argument (implicit from their behavior, never formally stated) seems to be that training on publicly available internet data is fair use under current law, and that settling would be an admission otherwise.
But Midjourney is also not inside ChatGPT, not in enterprise workflows the way OpenAI is, and has less to lose from prolonged litigation. OpenAI's stakes are much higher — they have millions of enterprise customers and governments evaluating their products. A clear, licensed content story is worth a lot more to them than winning an abstract fair-use argument.
What's Still Unresolved
The training data question. The Getty-OpenAI display deal sidesteps the fundamental question the courts are slowly working toward: does training a model on copyrighted images require a license? No deal can answer this. Only a verdict or legislation can. And no verdict has come.
Individual photographer compensation. Getty will receive fees from OpenAI. How much flows to the photographers who created the images in Getty's library is entirely unclear. Getty's royalty structure has always been controversial — contributors have historically complained about the split. Nothing in the announced deal addresses this.
What Stability AI does. Stability AI doesn't have OpenAI's resources or distribution. They're still in litigation with Getty and still facing questions about their business model. Their path to a licensing deal — if they want one — is much harder.
Whether Midjourney eventually folds. The combination of ongoing litigation, potential adverse rulings, and the industry normalizing licensing deals creates pressure on Midjourney to eventually negotiate. But "eventually" could mean years, and they've shown no signs of moving.
The Actual Takeaway
The AI copyright wars are not being resolved in courts. They're being resolved at negotiating tables.
Getty's deal with OpenAI is the clearest signal yet of how this ends: the major rightsholders will license to the major AI companies, take the money, call it "responsible AI development," and continue litigating against everyone else to drive up their own leverage at the next licensing table.
Individual artists — the ones who filed the original 2022 lawsuits — are not part of this negotiation. Their cases continue slowly through courts. Their images may or may not end up in licensed pools depending on whether they contribute to agencies that cut deals. The structural question of whether the creative ecosystem benefits from this shift, or whether the money concentrates at the agency layer (Getty, Shutterstock) and the AI layer (OpenAI, Google), remains wide open.
What's clear is that the old model — "scrape everything and fight in court if challenged" — is giving way to "license proactively and own the narrative." OpenAI learned this from watching what Stability AI's litigation exposure did to their credibility with enterprise customers. It was cheaper and smarter to be on Getty's side.
The question for everyone else in the image-generation space: are you Stability AI, or are you Shutterstock?
Related Reading
- Midjourney hardware announcement — June 2026 — what Midjourney is actually building while the lawsuits continue
- OpenAI GPT Image 2 — full breakdown — the image generation product this licensing deal directly improves
- Browse AI image tools — the full landscape of generation tools affected by these IP questions
Following the AI policy and copyright space? Subscribe to the ExplainX newsletter — we cover what's actually happening in AI, not just the press releases.