Copyright and AI in e-commerce: from product photo to ChatGPT output — who owns what?

“It’s on the internet, so we are allowed to use it.”

“The supplier sent the photo.”

“We paid the photographer.”

“AI created the image.”

“We only rewrote the competitor's text.”

These are all phrases that entrepreneurs use to assume that copyright is not an issue.

But in practice, product photos, advertisements, videos, product descriptions, designs and AI content in particular can lead to costly disputes.

Generative AI has not made this subject any easier.

On the contrary.

It has made it extremely simple to create, transform and reproduce content — whilst the legal question regarding rights is often only asked after publication.

Paying does not automatically mean that you own all rights

Imagine that you pay a photographer €3,000 for a complete photoshoot.

The invoice is paid.

The images are delivered.

Then the logical commercial thought is:

“These are our photos.”

But payment for an assignment and transfer of intellectual property rights are legally not the same thing.

The same applies to designers, video makers, copywriters, developers and marketing agencies.

That is why a professional collaboration should answer questions beforehand, such as:

Who is allowed to use the work?

For which purposes?

Can it be modified?

Can it be used internationally?

Can it be shared with partners or marketplaces?

Who owns any source files?

And what happens with AI usage?

These are not minor details if your brand later becomes worth millions.

“AI made it” does not solve the copyright problem

Generative AI can produce product images, ad copy, slogans, videos and designs in seconds.

But this raises at least two separate copyright questions.

The first question is about the input.

Which protected materials were used in or around the AI process?

The second is about the output.

Is the result sufficiently original? Does it resemble existing protected work too closely? And can the AI output itself be subject to copyright?

In 2025, the EUIPO published extensive research into precisely these two sides of generative AI: the use of protected content in model development and the legal questions surrounding generated outputs.

For businesses, that means primarily one thing:

Do not assume that “generated” means the same as “royalty-free”.

The legal status of pure AI output is not that simple

Within European copyright law, protection traditionally revolves around original work resulting from the creative choices of an author.

With fully autonomously generated output, a discussion therefore arises as to whether there is sufficient human creative input to justify copyright protection.

This does not mean that every AI image is automatically royalty-free.

But neither does it mean that a business automatically has the same exclusive position as with a work demonstrably created by a human author.

The legal analysis depends on the circumstances, including the human creative contribution and the way in which the result was achieved. The EUIPO explicitly mentions the protection of and risks surrounding AI output as a still-evolving part of the copyright landscape.

This is relevant for brands.

You do not want to discover only during a conflict that your main campaign image is harder to protect exclusively than you thought.

A prompt is no guarantee against infringement

Imagine an employee typing:

“Create a sports logo that strongly resembles the logo of brand X, but change enough so that we don’t have a copyright issue.”

That is not serious rights management.

The same goes for:

“Replicate this photo exactly, but change the person.”

Or:

“Rewrite this entire product description from our competitor so that no one notices it was copied.”

AI makes reproduction and transformation easier.

But it does not change the basic principle that protected creative expressions cannot be copied without limit.

And although a general style in itself is not the same as a concrete copyrighted work, a generated output can certainly contain elements of existing protected expression.

Therefore, commercial AI output must be assessed before publication, especially in brand campaigns and key assets.

Supplier product photos present a classic risk

Many webshops receive product photos via a supplier.

The supplier sends a Dropbox link.

The webshop publishes the images.

Everyone assumes everything is fine.

But do you know if that supplier took the photos themselves?

Or got them from the manufacturer?

Or copied them from another distributor?

And, more importantly: does your company actually have permission to use the photos for advertisements, marketplaces, social media and other commercial channels?

A reliable chain of title is just as relevant as a reliable product supply chain.

Influencer content is not automatically brand content

The same applies to influencers.

An influencer makes an excellent video.

Your company paid for the campaign.

The campaign ends.

Six months later, your marketing team uses the video again in Meta Ads.

Is that allowed?

That depends on the agreements made.

Was only an Instagram post agreed upon?

Also paid advertising use?

For how long?

In which countries?

Were images allowed to be edited?

Were AI tools allowed to generate new variants of the person or video?

Particularly with generative AI, these agreements are becoming more important.

Because one original photoshoot can nowadays be used to produce dozens of synthetic variations.

Contractually, it must then be clear what is and is not permitted.

Voices and faces also deserve attention

AI can create or manipulate realistic human images, voices and videos.

This affects not only copyright, but potentially also portrait rights, privacy, unfair commercial practices and the AI Act.

Since 2 August 2026, transparency obligations apply to, among other things, certain deepfakes and AI-manipulated content. The European Commission has recently published final guidelines and a voluntary code of practice for this.

A webshop working with a synthetic influencer must therefore look broader than:

“Does the ad look good?”

The relevant question becomes:

“What rights and transparency obligations lie behind this content?”

What happens to your own content if AI models want to use it?

AI and copyright are not just about what you get out of a model.

It is also about what others do with your content.

The European DSM Copyright Directive contains an exception for text and data mining of lawfully accessible works, but rightholders can expressly reserve their rights in certain situations, with machine-readable means being relevant for online content.

The AI Act has brought this subject further to the fore. Providers of general-purpose AI models now have copyright-related obligations, and the European Commission is working on technical mechanisms with which rightholders can indicate their TDM reservation.

For companies with valuable texts, photos, databases or creative catalogues, a new strategic question arises as a result:

Do you want your content to be usable for AI training, and if not, have you arranged that properly technically and legally?

This will become increasingly important in the coming years.

AI literacy is also necessary here

Marketing teams are often the fastest AI adopters within an organisation.

That makes sense.

The technology is fantastic for concepts, visuals, texts and variations.

But precisely these employees must understand that:

an output is not automatically royalty-free;

an input may not automatically be entered;

customer or supplier material may be confidential;

image rights can apply separately;

licensing terms of AI tools are relevant;

and publication always remains a separate commercial decision.

The AI Act now obliges professional providers and users of AI systems to take measures regarding the AI literacy of persons working with those systems on their behalf.

This makes “we just let everyone experiment” increasingly untenable as a company policy.

What to do in case of a copyright claim?

The other extreme is just as dangerous.

Entrepreneurs receive a letter stating:

“You have committed an infringement. Pay €8,500 within seven days.”

And pay immediately out of fear.

But a claim must be assessed.

Who is actually the rightholder?

On which work?

Which rights are alleged to have been infringed?

Was there perhaps permission or a licence?

How was the claimed amount calculated?

Was the material actually used as stated?

What correspondence exists?

Just because a claim is professionally drafted does not automatically mean that every legal or financial conclusion is correct.

Therefore, it is important not to impulsively acknowledge that an infringement has occurred before the file has been assessed.

AI makes evidence management more important

When content is created by multiple people, agencies and AI tools, the origin quickly becomes unclear.

Six months later, no one remembers:

which prompt was used;

which source images were entered;

which stock licence applied;

which freelancer created the original;

which AI tool was used;

which version finally went live.

Therefore, provenance — the origin of content — is increasingly important.

For major brands, content management is thus becoming almost a compliance process in itself.

Conclusion

AI makes content creation faster than ever.

But rights management does not automatically keep pace.

A professional e-commerce company must therefore not only know what it publishes.

It must also know:

where it comes from, who owns the rights and what is allowed to be done with it.

The companies that can scale AI most safely will not be the companies that avoid AI.

They will be the companies that know exactly where human control remains necessary.

Ecompliance

Does your business use photos, influencers, designers, external agencies or generative AI for commercial content?

Ecompliance can assess the chain of title, agreements and AI processes and help set up a structure that makes it clear who owns, uses and is allowed to publish what.