When news of the fashion company's deleted images reached the industry, reactions were swift. The focus immediately fell on the AI used to create the campaign.
Henrik Skagerlind Fasth, who works with the technology daily, believes the discussion has gone astray. According to him, it is not a new phenomenon in the fashion industry to lean on reference images to such an extent that they sometimes become copies. This is a pitfall that existed long before today's AI platforms broke through.
AI is not the problem with Filippa K's campaign. The lack of its own direction is, he says.
I think we miss the important point if we stop there. What happened at Filippa K could have happened entirely without AI. Starting from an existing image, drawing inspiration from the work of others, letting a reference motif steer too hard, that's nothing new in the fashion industry. Long before generative AI, art directors and photographers gathered image ideas from Pinterest and previous campaigns. The difference today is that the internet not only makes it easier to be inspired by others, but also easier to reveal when inspiration becomes a copy.
Even though AI tools have shortened lead times from sketch to finished production, they do not absolve the brand of ultimate responsibility. A human being is still needed to set the boundaries and steer the creative work.
To clarify the dynamics, he draws a direct parallel to traditional craftsmanship, where the client is always expected to deliver a clear brief.
AI did not make this choice. Someone did. The only difference is that the tool makes it faster and cheaper to go all the way from reference to finished image, without stopping to interpret in between. And that is where the real problem lies.
Think about how a brand works with a photographer. You don't hire a photographer and expect them to invent your brand for you. You come with a direction. You come with your own vision, an expression, a demarcation. That guidance is what makes the result yours.
He emphasizes that this applies regardless of whether production is done manually or digitally. If e-commerce companies input instructions that are based solely on the reference images of others, without adding their own idea, the result will inevitably reflect the lack of their own vision.
The tool does not give you an identity. It only amplifies the direction you already have, or the lack of it. If you take a reference image, turn it into a prompt more or less directly, change the clothes and think that's enough, then the result will be as watered down as the work behind it. It would look just as uninspired if a human had done it by hand.
When The System Takes The Helm
The Filippa K case is actually just an example of a broader phenomenon. Henrik Skagerlind Fasth sees companies in industry after industry starting to lean back and comfortably hand over all the thinking to the software.
We are seeing this everywhere with AI right now, not just in fashion. Companies take the technology for granted and let it replace thinking instead of sharpening it. AI is treated as the solution to the problem, when it is actually a tool you should direct towards the problem.
But despite the risks of going astray, his advice is not for e-commerce companies to slam on the brakes and stop using AI. The solution is instead to regain control and ensure that there is always a human being steering the process.
How: it's about reflection. Use AI where it actually solves a problem, where it removes friction, where it frees up time for what requires human judgment. But don't let the tool determine the direction. It is still your job to know what the brand is, what it is not, and where the line is between being inspired and copying.
The Role Of Suppliers In Development
Although the responsibility for a brand's expression ultimately rests with the company itself, the new technology is often marketed as a complete and worry-free solution for e-commerce companies. When Ehandel.se asks about the role that platform developers themselves play in the generation of plagiarism, Henrik replies:
First, to be honest: I may have put too much emphasis on the brands. AI tools are often sold as just fast and easy complete solutions. The word "end-to-end" is used frequently in the industry. And if you build and market a tool in that way, you also bear a responsibility for what it invites. You can't sell a shortcut and then put all the responsibility on the person who took it, he says and continues:
But I still want to be clear about where the line is. Ultimately, brands bear the responsibility for their own brands. No tool provider can take that away from them, and they shouldn't want to either. The tools' responsibility is different: to support the processes that brands already have, and to optimize them without compromising quality. Not to replace judgment, but to enhance it. And brands' responsibility is not to blindly rely on AI to solve every part of the chain.
To prevent copyrighted material from being published by mistake, he calls for more active controls. At his own company, they have chosen to develop a function that warns the user of similarities with existing works.
I think it is reasonable for more tools to have built-in barriers. At Lumoo, we have built a function specifically to identify and compare against existing image material, to catch this type of similarity before it reaches a campaign. We want it to be an active step in the flow, not something you discover only when the criticism comes. I'm not saying we've solved everything, but it's a conscious priority. As far as I know, we are quite alone in having it built in today, but I hope it will become the standard rather than the exception.
At the same time, the technology's format raises an inevitable question. If the systems now make it so extremely easy and cheap to transform a reference image into a finished campaign: doesn't the AI tools themselves drive exactly the shortcuts that cause the problems?
When something becomes extremely fast and cheap, the threshold for skipping the interpretation step is lowered. That is a real risk, and I don't want to pretend otherwise. But I don't think the answer is to slow down the tool artificially. Speed is also what makes the technology valuable. The answer is rather to build in friction in the right place. Not everywhere, but precisely where the shortcut risks becoming plagiarism. That is the difference between braking creativity and protecting it.
Ultimately, it is about shared responsibility, says Henrik. Tech companies need to take responsibility for how their systems are designed and what they invite, while e-commerce companies in turn must stop seeing software as a replacement for their own judgment.
In short: direction and judgment lie with the brand and can never be delegated to a tool. But the tool builders are not free either. Our job is to support and optimize the processes that already exist, design for the good outcome and not just the fast one, concludes Henrik Skagerlind Fasth.