The EU AI Act has created plenty of questions around labels, disclosures, and what counts as AI-generated content.
And those were exactly the kinds of questions we wanted to dig into when we sat down with media lawyer and AI expert Kelsey Farish.
We talked about the grey areas, what needs to be labeled, where disclosure might need to appear, and all the practical questions e-commerce teams are still trying to work through.
But some of the most useful points Kelsey raised were not necessarily the ones we had gone into the conversation expecting.
They were the second-order questions: what happens when consent becomes more important than the label itself? Could "human-made" eventually become its own selling point? And do brands need to start knowing much more about how an image was actually created?
That is probably why it helps to have a lawyer in the room.
Here are five points from the conversation that made us think about AI transparency a little differently.
1. Trust May Matter Before Enforcement Does
A lot of the conversation around regulation naturally focuses on penalties. And yes, the numbers can be significant.
But for many e-commerce brands, the more immediate risk may be much simpler: losing customer trust.
Kelsey pointed out that regulatory enforcement is only one part of the picture. If a brand presents itself as transparent around AI, but shoppers later discover that its content was generated or manipulated in ways they were not expecting, the reputational impact can arrive much faster than a regulator.
That makes AI transparency more than a compliance task.
It is also part of how a brand decides it wants to communicate with its customers.
2. The EU AI Act Is Not the Only Law That Matters
One of Kelsey's strongest points was also one of the easiest to overlook: just because an existing law does not mention AI does not mean it suddenly stops applying when AI is involved.
Copyright law still applies. Advertising rules still apply. Privacy and data protection still apply. Fraud, intellectual property, performer rights, and other existing legal frameworks do too.
As Kelsey explained, if someone uses AI to commit fraud, the fact that the fraud was carried out using AI does not make the Fraud Act irrelevant. The same principle applies elsewhere.
That is an important reminder for e-commerce teams because the EU AI Act can easily become the center of the conversation simply because it is new. But it is really only one layer of a much larger legal picture.
Using AI does not create a separate legal universe.
It adds another layer of questions to the rules that were already there.
3. Every Team Probably Needs Its Own AI Rules
One of Kelsey's most practical suggestions had very little to do with badges or labels.
Even if your company is not ready to publish a formal AI policy, your team should still know what its own rules are.
That could be as simple as deciding: What are we comfortable generating? What do we never alter without approval? When does a real person need to review the result? Which AI tools are acceptable for commercial work? Where do we disclose that AI was involved?
There may not be one perfect answer for every company, but having no answer at all becomes more difficult as AI spreads across everyday creative tools.
The important part is making those choices intentionally rather than discovering your AI policy one project at a time.
4. "Human-Made" Could Become a Label Too
One of the more interesting ideas from the conversation flipped the entire transparency debate around.
Instead of only labeling content that uses AI, what if brands eventually start identifying content that doesn't?
Kelsey mentioned a project in publishing that is exploring certification for human-authored work, effectively turning "human-made" into a positive signal.
It is easy to imagine a similar idea appearing elsewhere.
If AI becomes a normal part of commercial content production, then content made entirely by photographers, designers, writers, illustrators, or other creatives could become something brands actively choose to highlight.
Today, the unusual thing is AI-generated content.
A few years from now, the unusual thing may be content made without it.
5. The Biggest Shift May Be Knowing How Your Content Was Made
AI is increasingly invisible inside creative software.
A photographer may use one tool, a retoucher another, and a production platform several more. Somewhere along the way, AI might remove an object, extend a background, modify a person, regenerate part of a product, or create an entirely new scene.
That makes provenance much more important.
Brands may need to start knowing not only what asset they received, but how that asset got there.
Which parts were photographed? Which were generated? Which were altered? Was a real person involved? Was consent required? Was AI used only for cleanup or did it materially change the final image?
That level of visibility matters for compliance, but it also matters for consistency, approvals, and simply knowing what your customer is being shown.
And that may be one of the bigger changes AI brings to content production: the final image is no longer always enough information on its own.
A Final Word...
So those are probably some of the takeaways we walked away with from the conversation.
And, as Kelsey likes to remind everyone, she is a lawyer, not your lawyer. So if you want actual legal advice rather than what she jokingly calls "legal vibes," you can connect with her on LinkedIn.
Our takeaway from it all is to be intentional about how you use AI, understand what it is changing, and make sure the people responsible for the final content know what is happening along the way.
A lot of this is still being figured out, and that is probably not going to change overnight. But the more deliberate brands are now, the easier it will be to adapt as the rules and the norms around them become clearer.
Want the full conversation? Watch our webinar with Kelsey Farish, The EU AI Act and E-commerce.