If you have ever signed a modeling release and skimmed past the paragraph granting the brand a right to 'alter,' 'distort,' or 'composite' your image, this week's ruling out of New York should make you go back and read it again. A state court has declined to stop Rainbow USA from continuing to run AI-generated ad images of model Francheska Pujols, images she says depict poses, settings, and states of undress she never actually shot. The court did not decide the underlying question of whether Rainbow went too far. It simply declined to order the pictures taken down while the case proceeds. That distinction matters, and so does the reason Rainbow believes it is entitled to keep using them at all: an old contract clause never written with generative AI in mind.
The claim
Pujols, a New York City based model, entered a freelance modeling agreement with Rainbow in September 2024 for a photo shoot against a plain white background. According to her complaint, filed in New York State Supreme Court, the finished ads that ran after her contract lapsed placed her likeness in sexualized poses and settings she never posed for, including images described in court filings as showing her reclined suggestively on another woman's lap and seated with her legs spread on a barstool. The complaint, filed by professional fashion model Francheska Pujols in New York state court, alleges that Rainbow exceeded the scope of a license agreement that permitted the company to use specified photographs and videos depicting the model's likeness, and that Rainbow created and used AI-generated, AI-assisted, synthetic, derivative, composited or materially altered images depicting the plaintiff in poses, body positions, wardrobe presentations and settings that she never photographed or otherwise authorized. Rainbow disputes that any of this exceeds what it was permitted to do. In a memo filed on July 8, Rainbow frames the case as turning less on AI than on the terms of the parties' existing contract, asserting that the 2024 Freelance Model Waiver/Release/Indemnification Agreement grants it the right to broadly use Pujols' images "whether intact or in part, composite or distorted in character or form, cropped or altered, without restriction as to changes or transformations." On July 16, the court sided with Rainbow for now. Pujols' motion for a preliminary injunction was denied by the New York Supreme Court, in a suit alleging that Rainbow USA used artificial intelligence to create images of her in poses and settings she never appeared in, including at least one image exposing her undergarments. The disputed ads stay up while the case moves forward.
Why it matters
This case is not about scraping training data or an AI company's obligations. It is about something far more common and far more relevant to any independent designer, model, or creator who has ever signed a release: how far old, boilerplate contract language stretches when a brand runs it through a generative tool. Legal commentators tracking the case have been direct about what is actually being tested here. A recently filed lawsuit against Rainbow USA raises an increasingly important question for brands, agencies, retailers and content creators using generative AI tools: when does an authorized modification of a licensed image become an entirely new depiction of an individual requiring separate permission? That question did not exist when most standard releases were drafted. The allegations highlight an issue that many existing talent agreements were never drafted to address. While traditional licenses frequently permit parties to 'edit,' 'modify' or 'alter' photographs and videos, they rarely contemplate generative AI's ability to create entirely new visual depictions while maintaining an individual's recognizable face and likeness. Rainbow's contract language is not unusual. It is the kind of sweeping 'distort, alter, composite, without restriction' phrasing that has appeared in standard model releases for decades, written to cover cropping, color correction, and retouching. A generative model can now use that same language to justify manufacturing images the subject never posed for at all. Pujols herself has said this in sworn terms. In an affidavit, Pujols states that her contract only covered images taken during photo shoots and 'does not in any way authorize the creation of entirely new images, scenes, poses, or compositions that did not exist in the original content.' Whether a court ultimately agrees with her or with Rainbow will shape how every 'alter or modify' clause in the industry gets read going forward.
What to watch
The denial of a preliminary injunction is not a ruling on the merits. It only means Pujols could not show, at this early stage, that she is entitled to emergency relief. The underlying breach of contract, right of publicity, and related claims are still live, and Rainbow's dismissal arguments will need to be tested on a fuller record. Watch for the court's eventual ruling on whether 'distorted, altered, or composited' can be read to license wholesale synthetic recreation, since that holding will function as informal guidance for every brand and agency drafting or relying on similar language. Watch too for New York's own regulatory backdrop. The state's disclosure requirements for AI-generated performers in advertising add a separate compliance layer that exists independent of what any old release says. And watch how this case sits alongside the parallel dispute working through federal court in California, where influencer Molly Tranchin has sued the intimates brand EBY over an AI-altered video she says exceeded the one round of edits her contract actually permitted. That lawsuit raises novel questions about the use of generative AI in influencer partnerships and the extent to which existing agreements and emerging deepfake laws can police those practices. If you are negotiating a release right now, do not assume 'the usual language' still means what it used to. Spell out, in writing, that no AI-generated derivative, composite, or synthetic depiction may be created or published without your separate, specific approval. The Rainbow case is a preview of what happens when you do not.
Sources
- When Does an AI Alteration Become a New Depiction? Rainbow USA Lawsuit Highlights Risks for Talent Licenses and Synthetic Images
- Rainbow Says Existing Model Releases Already Cover AI-Generated Ads
- Rainbow Shops Can Use Disputed AI-Altered Model Images for Now
- Brands Quietly Exploit Right to Alter Influencer Content With AI
