The EU is about to make the invisible very loud. Under the new AI Act, the European Union is mandating that users be told when they are interacting with AI or viewing AI-generated or edited content. The goal is transparency, but the result will likely be a tidal wave of digital noise. As noted in a recent piece by Wired, this is leading to a legitimate fear of “disclosure fatigue.”
We have already seen this movie. It is the same energy as the cookie consent banners that now haunt every single website on the planet. We don’t read them; we just click “Accept All” as fast as humanly possible to get to the actual content. Who is this actually for? If the intent is to protect the user from being deceived, the solution is to flood them with so many warnings that the warnings themselves become a form of background radiation.
It is like putting a “contains air” label on a bag of chips. After the first ten bags, you stop seeing the label. After a hundred, you start wishing the label would just go away. If every AI-enhanced photo on Instagram and every LLM-assisted summary in a browser needs a tag, the tag ceases to be a warning and becomes a nuisance. (and we all know how that ends).
The friction here isn’t just psychological; it’s a UI nightmare. Developers now have to figure out how to shoehorn these disclosures into existing interfaces without ruining the user experience. Do you put a watermark on the image? A tooltip on the text? A pop-up that interrupts the flow? The result will be a cluttered, stuttering interface that serves the regulator’s checklist rather than the user’s needs.
The real irony is that this legislation treats AI as a discrete “event” or a specific “tool” that is added to a process. It assumes AI is a separate entity that can be neatly labeled. In reality, AI has become the plumbing of the modern web. It is not a feature; it is the infrastructure.
Trying to label every AI interaction is a fundamental misunderstanding of how software is built today. We are moving toward a world where the line between “human-authored” and “AI-assisted” is not just blurred, but non-existent. When a developer uses a copilot to write a function, then uses a linter to clean it, and then uses an LLM to document it, who is the author? Does the final code need a “Generated by AI” sticker?
It is a regulatory fantasy.
The EU is attempting to maintain a boundary that the technology has already erased. By forcing this transparency, they aren’t empowering users; they are simply forcing users to realize that they are living in a world where they can no longer trust their eyes or ears. That is a depressing realization to have via a mandatory pop-up window.
By Q1 2026, the EU will be forced to revise these disclosure requirements because the volume of labels will have rendered them functionally invisible to the average user. The regulators will realize that you cannot solve a systemic shift in computing with a labeling campaign.
The only thing this will actually achieve is a surge in “stealth” AI—tools that operate just below the threshold of what the EU considers “AI-generated” to avoid the paperwork. The tech will continue to evolve, the plumbing will get deeper, and the labels will just be more digital clutter for us to click away.
