If you work as a model, or you hire one, here is a new way an AI edit can go wrong. It is not just about consent to use someone's face. It is about what the software does to that face once it has permission to touch it at all. A menswear brand in Australia is now facing a racial discrimination complaint after a model says AI retouching lightened his skin and reshaped his nose. That is a different legal animal than the likeness disputes we have been tracking all year, and independent designers who lean on AI for campaign imagery should take note.
The claim
Elii Emeghebo, a Nigerian-Australian model, says Peter Jackson, a menswear company that specializes in suits, used AI to alter his skin tone and reshape his nose, among other changes to his facial features, according to a report from ABC News Australia. The company has not denied using AI. It says its creative team used "AI-assisted tools to create a substantially transformed image derived from the original campaign material," while categorically denying that race, color, or ethnic origin had anything to do with it. Paloma Cole, principal solicitor at the Young Workers Centre, represents Emeghebo. She says the alteration changed his image "in a way that removes a number of his racial characteristics," and that the company failed to pay him for the use of that altered image. A racial discrimination complaint has been brought to the Australian Human Rights Commission.
Why it matters
Most of the AI likeness disputes we cover turn on consent: did the model agree to have her face used this way, did the contract permit AI generation, was she paid for it. Those are real questions, and they are still live in the United States, where Francheska Pujols has twice sued Rainbow Shops over AI images she says she never posed for. But Emeghebo's complaint raises a separate question that contract language cannot fully answer. If an AI editing tool systematically lightens skin or reshapes features toward a narrower beauty standard, the harm is not only about authorization. It is about whether the output discriminates. That risk sits outside the right-of-publicity and copyright frameworks most brands rely on when they draft AI clauses into modeling agreements. A release that says the company may use "various technologies, tools, or production methods now known or later developed" does not immunize a brand if the technology's output has a discriminatory effect. For independent designers, the lesson is practical. If you are using AI to retouch or regenerate images of real models, you need to know what the tool is doing to protected characteristics, not just whether you have the rights to publish the result. Bias in training data is not a hypothetical concern raised by academics. It is showing up in actual campaign images, actual complaints, and now an actual filing with a human rights body.
What to watch
The Australian Human Rights Commission complaint is early. There is no ruling yet, and Peter Jackson disputes the discrimination claim while acknowledging the AI use. Watch how the Commission frames the legal question: whether an AI tool's output can constitute discriminatory conduct by the company that deployed it, even absent an intent to discriminate. In the United States, New York's Fashion Workers Act, in effect since June 19, 2025, already requires modeling agencies to get clear, conspicuous, and separate written consent before creating or altering a model's digital likeness through AI. That law was built to address consent gaps, not bias in the editing itself. If Emeghebo's complaint gains traction, expect advocacy groups and regulators on both sides of the Pacific to start asking whether disclosure and consent rules need a companion requirement: an accounting of what the AI actually changed, and why. For designers working with small budgets and AI tools that promise faster, cheaper retouching, that is worth building into your process now, before a complaint forces the issue.