Trademark law is emerging as a powerful new weapon in AI litigation, potentially posing greater risks to image-generation companies than copyright claims alone. While most legal attention has focused on whether AI models can train on copyrighted images, a quieter but equally significant battle is unfolding around brand protection, watermarks, and consumer confusion. The Getty Images v. Stability AI case demonstrates why this shift matters for everyone building or using generative AI tools like Stable Diffusion, Midjourney, and DALL-E.

What's the Difference Between Copyright and Trademark in AI Cases?

Most people assume copyright and trademark law protect the same thing, but they serve fundamentally different purposes. Copyright protects creative works themselves, such as photographs, illustrations, paintings, books, software, music, and movies. Trademark law, by contrast, protects consumers from confusion about where products or services come from. It safeguards company names, logos, slogans, product packaging, and distinctive branding.

This distinction becomes critical in AI litigation. Getty Images' watermark is not just text on a photograph; it functions as a source identifier that tells consumers where an image originated and who controls its licensing rights. When AI models generate images containing distorted or partial versions of recognizable Getty watermarks, the company argues this creates legal exposure under trademark law, even if copyright claims face hurdles.

How Can AI-Generated Watermarks Create Legal Problems?

Anyone who has experimented with image generators has likely noticed that AI occasionally produces garbled logos, partially formed text, incomplete watermarks, distorted branding, and nonsensical trademarks. Getty alleges that these outputs are not random artifacts but rather evidence that the AI model learned recognizable Getty branding during training. Because trademark law focuses on consumer perception, Getty contends that these outputs could falsely imply sponsorship, affiliation, endorsement, licensing, or origin.

Getty's U.S. trademark claims primarily arise under the Lanham Act, the principal federal trademark statute. The law provides several potential causes of action relevant to AI litigation:

• Trademark Infringement: Occurs when another party uses a protected mark in a manner likely to confuse consumers about source, sponsorship, affiliation, or approval. Unlike copyright law, courts ask whether consumers might mistakenly believe a product comes from or is associated with the trademark owner.
• False Designation of Origin: Prohibits marketing or presentation that falsely suggests a connection between parties. Getty alleges that AI-generated images displaying distorted Getty branding may create precisely this type of confusion.
• Trademark Dilution: Differs from ordinary infringement because it generally does not require proof of consumer confusion. Instead, owners of famous marks may claim that unauthorized uses weaken the distinctiveness of their brand or harm its reputation.

In April 2026, the U.S. District Court for the Northern District of California largely denied Stability AI's motion to dismiss Getty's complaint. The court concluded that Getty had plausibly alleged several claims sufficient to proceed into discovery, including trademark infringement, false designation of origin, trademark dilution, copyright infringement, and related California unfair competition claims.

Why Does the UK Case Matter Too?

Getty also pursued litigation against Stability AI in the High Court of England and Wales, producing important distinctions. The High Court rejected Getty's argument that the Stable Diffusion model itself constituted an infringing copy of Getty's photographs, concluding that the model does not literally store or reproduce copyrighted images in the manner Getty alleged. However, the court recognized limited trademark-related issues involving certain AI-generated outputs displaying Getty or iStock-style branding.

The court emphasized that these findings were narrow and fact-specific, not establishing that every AI-generated image containing distorted branding automatically violates trademark law. Nevertheless, the UK proceedings reinforced the growing importance of trademark theories in AI litigation.

Steps to Understand Trademark Risks in AI Development

• Recognize Watermarks as Trademarks: Many people think of watermarks as simply copyright notices, but they also function as branding, licensing information, source identification, and commercial goodwill. When AI reproduces recognizable portions of those identifiers, plaintiffs may argue trademark violations.
• Distinguish IP Claims: Different intellectual property claims contain different legal elements. A weakness in one statutory claim does not necessarily defeat others. Copyright and trademark claims can proceed independently, each with its own burden of proof and remedies.
• Monitor Training Data Transparency: As discovery proceeds in Getty v. Stability AI, the specific composition of training datasets will become increasingly scrutinized. AI developers should understand what branded content may be present in their training materials and how models might reproduce recognizable branding.

The Getty Images v. Stability AI case remains in early stages, with discovery, expert testimony, summary judgment, and possibly trial ahead. However, the court's April 2026 ruling allowing trademark claims to proceed signals that this legal theory has real teeth. For AI companies, artists, photographers, and technology firms, the message is clear: trademark law may pose risks that are just as significant as copyright concerns, and the outcome of this litigation could significantly influence how AI developers build future models.