Using AI-generated images commercially: the three things that decide it

The more useful question is: six months from now, can you show where this image came from, who approved it, and what you are allowed to do with it?

Three things decide that. None of them is your prompt.

This is not legal advice, and we are not lawyers. It is what we have learned producing commercial visuals with AI, and what we would want a marketing team to have sorted out before they scale.

1. The terms of the tool you used

Every generator has its own rules, and they differ more than people assume. Before you build a production process on a tool, read its terms for four things:

Commercial use. Is it allowed at all, and does it depend on your plan? Some tools allow it only on paid tiers, and the free tier you tested with is not the tier you are now running on.

Exclusivity. Can anyone else generate something very similar? With most tools, yes. That matters if you were planning to use the output as a brand asset rather than as a product shot.

What happens to your uploads. If you upload your own product photography as reference, is that material used to train the model? For an unreleased product, this is the question that should stop you.

Indemnity. Some providers offer to defend you if a claim is made about output; most do not. Know which one you are on.

This is the part teams skip, and it is the only one of the three you can settle completely in an afternoon.

2. Who owns the result, where you sell

Copyright protection generally depends on human creative input. A purely generated image, made by typing a prompt, sits awkwardly against that idea, and the answer is not identical in every country you sell into.

The practical consequences matter more than the doctrine:

You can almost certainly use the image. That is rarely the problem.

You may not be able to stop someone else using something nearly identical. If an image is doing brand work, not just product work, that is a real exposure.

The more human direction, selection and editing went into it, the stronger your position. This is one of several reasons we keep a person in the loop rather than generating straight to publication.

If it matters commercially, ask your own counsel about the markets you sell in. One answer for the whole world does not exist here.

3. What you have to disclose

This one has a date on it. The transparency obligations in Article 50 of the EU AI Act apply from 2 August 2026 to AI-generated content (source: EU AI Act, Article 50). If you sell into the European Union, that applies to you whether or not you are based there.

What this means in practice is less dramatic than the headlines suggest, and it is mostly an operations question: you need to know which of your assets were generated, and you need to be able to say so. If you cannot answer that today for the images already on your website, that is the gap to close first.

What to actually do

Four things, in this order. None of them requires new software.

Record provenance per asset. For every image: generated or photographed, which tool, what reference material, who approved it, and for which markets and channels it is cleared. Add the field now, while your library is the smallest it will ever be again.

Put a person on approval. Not on everything, but on everything that leaves the building. The failure mode with generated imagery is not a spectacular mistake, it is a plausible one: a product in a colour you do not sell, a detail that does not exist on the real thing, a room that is subtly the wrong country.

Be careful with people and places. Recognisable faces, real locations, other people's products in the background and anything that resembles an existing campaign are where generated images create problems that have nothing to do with AI law.

Tell your partners what they may do. If dealers and retailers republish your images, your rules travel with the files, or they do not travel at all. Put the usage terms where the images are, not in an email nobody kept.

Why we work the way we do

We produce commercial visuals for brands, so these questions are ours before they are yours. Three choices follow from that.

A person from our team is in the loop on what we deliver, which is both a quality decision and the reason there is human judgement in the work. Visuals are produced against the Brand DNA of your brand, so output stays inside your own rules rather than drifting toward whatever the model finds attractive. And a commercial usage licence is part of LensLab Studio, so the permission question is answered before you publish rather than after.

You can see how we work, or look at our use cases for what that produces.

In closing

The legal picture around AI images will keep moving. Your position is better if your house is in order regardless of which way it moves: provenance recorded, approval by a person, clear terms with the tools you use and with the partners who republish your work.

That is work you can do this quarter, and it is the same work that makes producing a range at scale manageable in the first place.

Want to see what this looks like with a creator in the loop? Try LensLab free with five credits.