While You Create.com

Safety Published August 1, 2026 ·2 min read

Two AI uses become prohibited in the EU on 2 December 2026

The European AI Act gains two prohibitions covering manipulated imagery of identifiable people. What the text actually says, and what changes for a family, and what does not.

Cover of While You ScrollThe book behind this articleWhile You Scrollfor teenagers (13 to 18)

On 2 December 2026, two new prohibitions in the European AI Act start to apply. Small in extent, large in reach, and worth reading calmly, because exaggerated versions are circulating in both directions.

What becomes prohibited

Article 5, the list of prohibited practices, gains two new points. The first prohibits AI systems that generate or manipulate realistic image, video or audio content depicting an identifiable person in an intimate way, without that person's explicit consent. The second prohibits systems aimed at generating child sexual abuse material.

Both sit on the regulation's highest penalty tier, the same one as social scoring: up to EUR 35 million or 7 per cent of worldwide annual turnover.

What it does not prohibit

This is the part that almost always gets lost. The regulation adds two paragraphs that narrow the scope. Placing a system on the market is prohibited only where that generation is its intended purpose, or where it is a reasonably foreseeable and reproducible outcome and the system lacks reasonable technical safeguards. And using one is prohibited only where the deployer uses it for that purpose.

In other words: this is not a general ban on image generators. It targets applications built for this, and those that allow it with no barrier at all.

What changes for a family

On the supply side, quite a lot. Applications advertised for precisely this can no longer be offered or used in the EU, and they move into the category authorities pursue with the heaviest means.

And on the phone at home

Less, and it is worth being honest about that. A European rule does not erase what is already installed or what gets passed around in messaging apps. It removes the offering from the market and gives a clear basis for reporting something. It does not stop it appearing.

The word that works at home

Consent. The rule speaks neither of taste nor of intent, but of a freely given, specific, informed, unambiguous and explicit authorisation from the person depicted. That is a standard a teenager grasps without reading the regulation, and it applies to a class photo exactly as it does to a composite: the question is not whether it is funny, but whether the person in it agreed.

That conversation is more use than the prohibition, because it arrives earlier. The law acts once something exists. The question about consent acts while somebody still has their hand over the button.

Sources

FAQ

When does it apply?

From 2 December 2026. The two new prohibitions carry their own date inside the regulation, later than the rest of the chapter.

Does it ban image generators in general?

No. The regulation narrows the scope explicitly: it covers systems whose intended purpose is this, or which foreseeably allow it without reasonable safeguards, and deployers who use them for that purpose.

Is it any use if the app sits outside the EU?

It removes the offering from the European market and gives a report a legal basis. It does not make what already circulates disappear, and it is better not to promise that.

The three books in the series

One approach, create instead of consume, in three versions depending on who is reading.

for teenagers (13 to 18)

While You Scroll

This book

The whole series on WhileYouCreate.com

The site has the free companion pages for all three books, the downloadable material and the journal.

Go to WhileYouCreate.com →Learn AI as a family →