The imagery your brand has published since its founding — every runway photograph, every lookbook, every editorial feature, every campaign — has almost certainly been incorporated into AI training datasets. LAION-5B, the training dataset underlying Stable Diffusion and several other major AI image models, contains approximately 5.85 billion image-text pairs scraped from the public internet. Fashion imagery is significantly represented.
This happened without notification, without consent, and without compensation. Whether it was legal is currently being decided in courts in the US, UK, and EU. Whether you can do anything about what has already happened is a harder question. What you can do going forward is the practical focus of this piece.
The live litigation: what the cases are arguing
Several major cases are currently in active litigation that will determine the legal status of AI training on copyrighted imagery. The cases most directly relevant to fashion designers:
Andersen v. Stability AI, Midjourney, DeviantArt (N.D. Cal.) — Filed by a group of visual artists whose work appears in AI training datasets. The claims include direct copyright infringement (for the training itself), DMCA violations (for stripping copyright management information), and right of publicity claims. Currently proceeding after surviving a motion to dismiss on key claims.
Getty Images v. Stability AI (D. Del.) — Filed by Getty Images alleging Stability AI used approximately 12 million Getty images without licence to train Stable Diffusion. Getty is asserting copyright infringement, DMCA violations, and trademark infringement (for AI output that replicates Getty's watermarks). A settlement process was announced in early 2025.
Various European cases — EU courts are also examining these questions under different legal frameworks, including GDPR (for imagery containing identifiable persons) and the EU copyright directive's text and data mining exceptions. EU outcomes may diverge from US outcomes.
The central legal question in most US cases: does training an AI model on copyrighted images without permission constitute copyright infringement? The defendants' primary argument is fair use. The plaintiffs' primary argument is that training is not transformative and creates competing products that harm the market for the original work.
You cannot un-train an AI model on your past work. But the legal question of whether that training was infringement — and whether compensation is owed — is actively being litigated, and the outcome will have significant consequences for the fashion industry's relationship with AI tools.
What the Getty settlement means for fashion brands
The Getty-Stability AI settlement, announced in early 2025, is significant not for its specific terms but for what it represents: the first major commercial resolution of AI training data disputes. The settlement framework establishes a licensing model for AI training on commercial photography archives. While the specific terms are confidential, industry reporting from Business of Fashion and The Fashion Law suggests the framework involves retrospective licensing payments and prospective licensing arrangements for future training.
The implications for fashion brands: if you have substantial photography archives — extensive runway documentation, large campaign libraries, significant editorial archives — they may have commercial value in AI licensing contexts that didn't exist two years ago.
The output infringement question
Separate from the training data question is the question of whether AI-generated output that closely resembles copyrighted training material itself infringes that copyright. This is particularly relevant for fashion brands with distinctive visual identities.
If a competitor uses an AI tool to generate designs that closely replicate your distinctive aesthetic, patterns, or visual identity — and that AI tool was trained on your published imagery — you may have both copyright and trade dress claims. Several fashion brands are actively monitoring AI-generated output for similarities to their IP and building documentation for potential claims.
What you can do right now
Register your key copyrights. Copyright registration is required to sue for statutory damages in the US. Registration is inexpensive (the Copyright Office's basic filing fee is $45-$65 per work), creates a timestamped public record, and significantly strengthens your legal position in any infringement dispute.
Implement technical opt-out measures. The emerging standard for technical AI training opt-out is the robots.txt protocol combined with the Spawning.ai opt-out API and similar services. While these technical measures don't affect what's already in training datasets, they signal your intent going forward. Update your robots.txt to include AI training crawlers:
User-agent: CCBot
Disallow: /
User-agent: GPTBot
Disallow: /
User-agent: anthropic-ai
Disallow: /
Note: the legal status of these opt-outs is not definitively established. They are a reasonable protective measure, not a guaranteed legal shield.
Document the originality of your work. In any future infringement dispute, you'll need to demonstrate that your work reflects original human creative expression. Document your design development process — sketches, mood boards, material explorations, the creative reasoning behind key decisions.
Monitor for visual similarity in AI output. Tools for detecting AI-generated imagery that closely resembles specific training material are developing rapidly. For brands with distinctive visual identities, periodic monitoring of major AI image generation platforms for outputs that closely replicate your aesthetic is worthwhile.
Include AI training restrictions in your photography contracts. New photography and content creation agreements should include explicit prohibitions on the use of created imagery for AI training purposes, representations that the photographer has not submitted your imagery to AI training datasets, and IP ownership provisions that make clear the brand's ownership of all imagery rights including AI-related rights.