What Happens to an Image When Ambiguity Ends?
.jpg)
For a while, ambiguity was enough.
An image could circulate in the gap between fascination and doubt.
Was it photographed, generated, or simply left unclear on purpose?
From 2 August 2026, that question may become harder to avoid in Europe.
This is when the transparency obligations in Article 50 of the EU AI Act are due to start applying. That does not mean every image involving AI suddenly becomes questionable. Many ordinary editing tools already rely on AI in one way or another.
This is not about every image touched by AI. It is mainly about AI-generated images, especially when they are presented in a way that could be taken for photographs or real scenes.
For some of that content, disclosure will no longer depend only on ethics or platform policy. In some cases, it will be a legal requirement.
The broader logic is clear as well: some AI-generated content is expected to carry a machine-readable, detectable trace, so that platforms and other systems can identify it more easily as artificially generated or manipulated.
That does not mean perfect traceability. Metadata may not survive every platform or workflow.
But that is not the part that interests me most.
For a while, part of the force of certain AI images came from uncertainty.
Not only from what they showed, but from the hesitation they produced.
Could they pass as photographed?
Could they remain unclear long enough to keep their effect?
If that ambiguity becomes harder to maintain, even imperfectly, the image has less to lean on.
Less on hesitation.
Less on the old game of is it or isn’t it?
Then another question matters more. Not only how it was made, but why it exists at all. And whether there is still a point of view behind it.
Transparency will not settle everything.
It will not create trust by itself. But it may clear part of the haze.
After that, the image has to stand a little more on its own.
That is only one part of the story.
The question of how these systems were trained, and on whose images, is another one entirely.
For a more accessible overview, the European Commission’s FAQ on transparent AI systems and its page on the Code of Practice on marking and labelling of AI-generated content are a good place to start.
The photographs are gathered here:
Selected Works
.jpg)
Another very good article Philippe. To me an AI image should never be allowed to be called a photograph. I also wish that photography social media would ban AI images or at the least have a separate section for them. Like you I use Vero and I was astonished and annoyed a couple of weeks ago that a well respected hub was featuring a obviously AI image as its "photo of the day".
ReplyDeleteThank you. I understand that reaction. For me, the key point is not to deny that AI images can exist, but to make sure viewers know what kind of image they are looking at.
Delete