If you model for a living, or you hire models for your own label, the ruling that came out of the New York Supreme Court on July 16 deserves your full attention. A judge declined to stop Rainbow Shops from continuing to publish AI-altered images of model Francheska Pujols, images she says she never posed for and never approved. The case is not over. But the order tells you something you need to know now, before your next shoot: a broadly worded release, signed long before anyone was thinking about generative AI, may already give a brand more room to manipulate your image than you think.

The claim

Pujols filed suit in New York state court, alleging that Rainbow exceeded the scope of a September 2024 modeling release when it used AI to generate new images built from photographs she had posed for and approved. Her complaint says Rainbow kept her recognizable face while digitally altering her hair, clothing, body positioning, backgrounds, poses, and overall composition, producing images she never sat for. She asked the court for a preliminary injunction to pull the images down while the case proceeds. On July 16, the court said no, leaving the images up for now. Rainbow's position, laid out in a July 8 memo, is that the fight is not really about AI at all. The company points to language in the original release granting it the right to use Pujols's images, in the company's words, whether intact or in part, composite or distorted in character or form, cropped or altered, without restriction as to changes or transformations. Rainbow also says the release waived Pujols's right to inspect or approve finished advertising and released the company from liability for blurring, distortion, alteration, optical illusion, or use in composite form. Read that way, an AI-generated derivative is just another kind of alteration, one the model already agreed to.

Why it matters

This is the fight every independent model and every brand that hires models should be watching. Most standard releases were drafted for a world of retouching, cropping, and color correction, not for tools that can place a recognizable face into a scene the model never occupied. Rainbow's argument asks the court to read old contract boilerplate as if it already anticipated generative AI, and the July 16 order, denying only the preliminary injunction rather than ruling on the merits, does not settle whether that reading holds. It simply means Pujols could not clear the high bar required to force the images offline before trial. For designers who work with independent models, the practical lesson is not about this defendant's guilt or innocence. It is about your own paperwork. If your release contains sweeping alteration language borrowed from an old template, a court may someday read it exactly the way Rainbow is reading its own. New York's Fashion Workers Act, which took effect June 19, 2025, was written with this problem in mind. It requires clients to obtain models' written consent before creating or using digital replicas of their likeness, a protection layered on top of, not a replacement for, whatever the underlying modeling contract says. Whether that consent requirement reaches back to cover a September 2024 release signed before the Act's effective date is exactly the kind of question this case, or the next one like it, will have to answer.

What to watch

Watch for Rainbow's answer or any motion to dismiss following its July 8 memo, and watch whether Pujols's team seeks an interlocutory appeal of the injunction denial. Watch, too, for how the court eventually treats the interplay between old-style broad alteration clauses and the Fashion Workers Act's newer, AI-specific consent requirement, since that tension sits at the heart of the case. In the meantime, if you are a working model, read your release before you sign it and ask whether the word alteration was ever meant to cover a wholly new, AI-generated pose. If you are a designer or brand hiring models, do not assume your existing template protects you. Draft explicit AI and digital-replica language, require written consent for any synthetic derivative, and give your models real approval rights over anything AI touches. The law here is still being written case by case, and right now, the party with the more carefully drafted contract is winning.

Sources