AI lawsuits
AI-image lawsuits are no longer hypothetical.
They hit brands first.
When a campaign features an AI-generated face that resembles a real person, the chain of liability runs in one direction: not toward the model provider, but toward whoever paid for and published the image. That's the brand. That's the agency. The model vendor has terms-of-service indemnification carve-outs; you have a campaign to defend.
Who actually gets sued
Plaintiffs target the deepest pocket on the publication side. That is the advertiser or the agency of record. Foundation-model providers are typically shielded by terms of service that disclaim downstream liability and by jurisdictional distance. Even when their training data is the root cause, recovering against them is procedurally difficult. The brand is faster and easier, and visibly using the contested image.
What the damages look like
Settlements for unauthorized-likeness cases typically combine a withdrawal of the campaign, a public correction, and damages calibrated to the campaign reach. In Germany, courts apply a "lizenzanaloge Schadensberechnung", compensation calibrated to what a license would have cost, plus a punitive multiplier for unauthorized use. In the US, statutory damages under right-of-publicity statutes can stack with state-by-state claims. Reach matters: an OOH campaign with national exposure is not a small case.
Why "we used AI" is not a defense
Personality rights protect identifiable likeness regardless of medium. The fact that an image was generated rather than photographed does not break the protection. It shifts the analytical question from "is this her photograph" to "is this recognizably her." Plaintiffs and courts have shown they can answer the second question, and resemblance is increasingly a sufficient predicate.
Licensing as risk-shifting
A licensed identity assigns the rights chain: the depicted person consented, on documented terms, to specific commercial use. The license bounds the scope and creates an evidentiary trail. If a complaint arrives, you can answer it with a contract. The case-law calculus shifts from "did you have rights" (an open question) to "did you stay within scope" (a contract question). That is a far more defensible position.
Every AI-generated face is a legal risk. Unless it is licensed.
FAQ
Frequently asked.
Has anyone actually been sued over an AI-generated image?+
Yes. Right-of-publicity and personality-rights claims have been filed in multiple jurisdictions against advertisers and platforms using AI-generated likenesses. The litigation pattern is established; what is still moving is damages calibration.
Does buying AI image generation from a vendor protect me?+
No. The vendor terms of service almost always carve out downstream liability. The vendor sells you tools, not a rights chain. The campaign you publish is yours to defend.
Can I be sued if the face only resembles a real person?+
Yes. The legal test is recognizability, not exactness. If a reasonable observer would identify the depicted person, the claim is viable. Plaintiffs have prevailed on resemblance alone in multiple jurisdictions.
How do licensed identities reduce exposure?+
They convert open-ended likeness risk into a defined contract. The depicted person has consented to documented commercial use. Disputes become contract questions (was the use within scope) instead of rights questions (did you have rights at all).
Ready?
Get early access to FaceLedger for companies. Configure licensed identities directly into your generative-AI workflow.
Request agency access