On August 4, 2026, professional models Taber Maier and Ariel Howard sued Omni Hotels Management Corporation in Los Angeles County Superior Court. Their complaint says Omni took photographs from a single licensed photoshoot and used artificial intelligence to build a Christmas advertisement they never posed for, never approved, and say they would never have agreed to shoot. As of August 16, no news outlet has covered the case. This article works from the complaint itself.
The dispute fits in one sentence. The models licensed one defined digital campaign for three years, and Omni allegedly answered their objections by saying that “minor alterations to the images are covered by the license.” The complaint says the alteration changed their facial features, their hairstyles, and the entire scene around them. One side calls that a minor edit. The other side calls it a fabricated depiction. Between those two positions sits the question every brand running generative editing tools now has to answer: when does an AI edit stop being an edit and become a new use of a person?
The case lands three months after a nearly identical fight started in New York, where model Francheska Pujols sued the retailer behind Rainbow Shops over AI-altered campaign images. Two suits, two coasts, the same fault line. The pattern is no longer a curiosity. It is a docket category, one we have been mapping since our audit of AI litigation in state courts.
What the complaint says happened
According to the complaint, Maier and Howard were booked on March 21, 2025 for a campaign shoot at the Omni San Diego Hotel at 675 L Street. The booking paid each model a day fee plus a separate usage fee. The usage grant was narrow and specific: one defined digital campaign, three years, no broadcast, no print.
The complaint adds a second layer of consent that did not exist in model bookings five years ago. Paragraph 14 alleges that, as a condition of the booking, the models refused to allow their names, images, voices, likenesses, or performances to be “used, replicated, or modified by or for any artificial intelligence systems, tools, or applications” without prior written consent. Neither model gave that consent, the complaint says. The fact that a March 2025 hotel campaign booking carried an explicit no-AI condition is itself a data point about where the modeling industry has moved since SAG-AFTRA made digital replicas a strike issue in 2023.
Roughly six months after the shoot, the models allegedly discovered a published Omni advertisement in which their likenesses had been, in the complaint’s words, “materially altered by artificial intelligence” into a Christmas or holiday-themed ad. Paragraph 16 is the complaint’s technical core: the alteration “was not a standard crop, color correction, or ordinary advertising edit.” On information and belief, it changed the models’ facial features, their hairstyles, and the surrounding scene, placing them in “a fabricated commercial holiday-themed depiction that they never sat for, approved, or authorized.”
After the models and others objected, the complaint says, Omni kept running the ad for a period of time before pulling it, and only after repeated requests. When confronted, Omni allegedly took the position that minor alterations were covered by the license while producing no document that authorized AI modification. Omni has not yet answered the complaint, and that quoted position comes from the plaintiffs’ filing, not from anything Omni has said in court.
The complaint pleads two causes of action: common law misappropriation of likeness and violation of California Civil Code section 3344, the state’s statutory right of publicity. It demands a jury, seeks actual damages including emotional distress, punitive damages under Civil Code section 3294, disgorgement of profits, a constructive trust over money attributable to the use, and attorney’s fees under section 3344(a).
What generative editing actually does to a photograph
The legal fight will be argued in contract language, but it turns on a technical fact about how modern image editing works. A conventional retouch, the kind license clauses were written for, is arithmetic on existing pixels. Cropping discards pixels. Color correction remaps their values. Even aggressive airbrushing displaces and blends pixels that the camera captured. The photograph remains the source of truth, and every output pixel traces back to a captured one.
Generative editing breaks that chain of custody. Tools like Adobe’s Firefly-powered Generative Fill, and the inpainting modes in diffusion systems such as Stable Diffusion and Flux, work by masking a region of the image and synthesizing replacement content from a trained model. Inside the mask, the pixels are not adjusted. They are generated: sampled from a probability distribution the model learned across billions of training images, conditioned on a text prompt and on the pixels surrounding the mask. Swap a hotel lobby for a Christmas scene and the fireplace, the garlands, and the lighting on the subjects’ faces are statistical inventions, not captures.
That distinction matters most when the mask touches the person. Relighting a face for a new synthetic scene, adjusting a hairstyle to fit a seasonal look, or regenerating a jawline that the compositing damaged all mean the model is synthesizing the subject’s biometric geometry rather than reproducing it. The output resembles the person because the system was conditioned on their photograph, but the specific face in the ad is a rendering, the same way a portrait painter’s output resembles but is not the sitter. If the complaint’s allegations are accurate, the holiday ad contains faces that no camera ever recorded.
This is why “edit” is the wrong mental model and why the license fight is genuinely hard. A clause permitting “alterations” was drafted against a technology where alteration meant transformation of the captured image. Diffusion inpainting is closer to commissioning a new image with the old one as reference. Contract language has not caught up, and the two 2026 lawsuits are the collision.
The license question, and the Rainbow parallel
Omni’s alleged defense has already been road-tested in another courtroom. In Pujols v. Rainbow USA, filed in New York Supreme Court on May 22, 2026, withdrawn for settlement talks, and refiled on June 15, model Francheska Pujols alleged that a September 2024 license allowing “limited modifications such as cropping, formatting, or stylistic alterations” of her campaign photos did not authorize AI-generated images showing her in poses and settings she never shot. Rainbow’s motion to dismiss argues the opposite: the contract granted rights to alter, transform, and composite her likeness, ordinary contract interpretation resolves the case, and no novel AI doctrine is needed. In July, a Manhattan judge denied Pujols’ preliminary injunction, leaving the disputed images up while the case proceeds, as Bloomberg Law reported.
The Omni complaint sets up the same collision under California law, with two differences that favor the plaintiffs on paper. First, the alleged license here was unusually narrow: one defined digital campaign, three years, no broadcast, no print, plus an express no-AI condition. The narrower the grant, the harder it is to read a fabricated holiday scene into it. Second, California Civil Code section 3344 is a sharper weapon than New York’s privacy statute. It awards the greater of $750 or actual damages, plus profits attributable to the unauthorized use, with a burden shift the Court of Appeal confirmed in Olive v. General Nutrition Centers: the plaintiff proves gross revenue attributable to the use, and the defendant must prove its deductible expenses. Punitive damages and prevailing-party fees stack on top.
Note what section 3344 does not ask. The statute punishes anyone who “knowingly uses another’s name, voice, signature, photograph, or likeness” for advertising without consent. The word likeness does not care whether an airbrush or a diffusion model produced the depiction. The AI question in this case is not whether the statute applies. It is whether the license excuses the use, and whether the manner of alteration proves the use was knowing. A company that changes a model’s face and surroundings by prompt will struggle to claim it did not know it had left the territory of the photograph it licensed. We covered the same structural move, decades-old statutes absorbing brand-new machinery, in our analysis of wiretap law applied to AI phone agents.
California’s legislature has also been paving this road. AB 2602, effective January 1, 2025 and in force before the March 21, 2025 shoot, added Labor Code section 927, which voids contract provisions allowing digital replicas of a performer’s voice or likeness in place of work they would have performed in person unless the provision contains a reasonably specific description of the intended uses and the performer had union or legal representation. The complaint does not plead section 927, but the statute defines the state’s policy direction: consent to synthetic use must be specific, not inferred from general alteration language. And in late 2025, SB 683 amended section 3344 to expressly authorize injunctive relief, strengthening the remedy set for exactly this category of dispute.
What the complaint does not establish
Fair reading requires listing what the filing leaves open, because several of its central technical claims are pleaded on information and belief rather than demonstrated.
The complaint never explains how the plaintiffs know artificial intelligence altered the images. Paragraph 16’s description of changed facial features, hairstyle, and scene is consistent with diffusion inpainting, but it is also consistent, at least in principle, with an old-fashioned composite assembled by a skilled retoucher. The complaint names no tool, no vendor, and no forensic basis for the AI attribution. That does not weaken the license-scope theory much, since a hand-built fabricated holiday scene would exceed a narrow campaign license just as thoroughly. But it matters for the no-AI booking condition in paragraph 14 and for the punitive damages narrative, both of which assume AI specifically was used.
The proof, if it exists, lives in Omni’s production pipeline, which is where discovery will go. Layered working files look completely different from flattened diffusion outputs. Agency invoices name the tools billed. Adobe’s Firefly attaches C2PA Content Credentials to generative outputs by default, though that metadata rarely survives export and its absence proves nothing. Prompt logs, asset management timestamps, and vendor emails will settle the question one way or the other. The complaint is a claim about what those files will show.
Omni has not answered, and no hearing has occurred. The “minor alterations” quote attributed to Omni is the plaintiffs’ account of a pre-suit exchange. A plausible defense reading exists: if the booking paperwork contained standard retouching or alteration language, Omni will argue, as Rainbow has, that this is a routine contract interpretation case, and the Pujols injunction denial shows at least one court declining to treat AI alteration as self-evidently out of bounds. Everything in this article describing the parties’ conduct is allegation, not established fact.
What happens next
Omni’s response is due on the standard California timeline, roughly thirty days after service, so an answer or demurrer should appear by early fall. A demurrer would test whether the complaint’s allegations, taken as true, state a claim. Given how squarely the facts alleged fit section 3344’s elements, the more likely battleground is summary judgment after discovery establishes what the license actually said and what the production files actually contain.
Three things are worth watching beyond this docket. First, whether discovery surfaces the production artifacts, because a documented diffusion workflow inside a major hotel brand’s ad pipeline would be the clearest public evidence yet of how routine generative editing has become in commercial photography. Second, whether the case settles quickly, which is how most section 3344 claims against brand defendants end, and which would leave the license-scope question unanswered for the next plaintiff. Third, how the Pujols court rules on Rainbow’s dismissal motion, since a New York holding that “alter, transform, and composite” covers AI generation would hand every brand defendant a template, while a denial would do the opposite.
The durable question outlasts both cases. A photograph license grants rights in an image. Section 3344 protects the person. Generative editing splits those two things apart, because the output ad is no longer the licensed image but still depicts the licensed person. Until booking contracts name synthetic use explicitly, the way paragraph 14 of this complaint says these models already demanded, that gap is where the lawsuits will keep coming from.
Frequently asked questions
What does the lawsuit accuse Omni Hotels of doing?
The complaint alleges Omni took images from a March 2025 campaign shoot licensed for one digital campaign and used artificial intelligence to alter the models’ faces, hairstyles, and surroundings into a holiday advertisement they never posed for, then published it without consent. Omni has not yet responded in court.
Does California law treat AI alterations differently from Photoshop edits?
Civil Code section 3344 does not distinguish by tool. It prohibits knowing commercial use of a person’s likeness without consent, however the depiction was made. The tool matters instead to license scope, since a clause permitting alterations of a photograph may not cover synthesizing a new depiction, and to whether the use was knowing.
What damages can models recover under Civil Code 3344?
The greater of $750 or actual damages, plus profits attributable to the unauthorized use that are not already counted, with the plaintiff proving gross revenue and the defendant proving deductible expenses. Punitive damages and prevailing-party attorney’s fees are also available, and since a late 2025 amendment the statute expressly authorizes injunctions.
Has Omni responded to the lawsuit?
No. As of August 16, 2026, no answer or motion appears on the docket, and no news organization has reported on the case. The only account of Omni’s position is the complaint’s allegation that the company called the changes minor alterations covered by the license.
Methodology: This analysis is based on the complaint in Maier and Howard v. Omni Hotels Management Corporation, filed August 4, 2026 in Los Angeles County Superior Court and retrieved on August 16, 2026 from a commercial index of state trial court dockets, together with the statutory texts of California Civil Code sections 3344 and 3294, Labor Code section 927, and court records and reporting in Pujols v. Rainbow USA. All descriptions of the parties’ conduct are allegations from the complaint, which Omni has not yet answered. Nothing here is legal advice. This article is part of our series on AI litigation in state courts, which began with an audit of how AI lawsuits are counted.