Amsterdam court bars xAI and Grok from creating "undressed" images without consent

An Amsterdam court has barred xAI from generating and distributing images of people without clothes in the Netherlands without their consent. The ruling directly affects Grok and shows that the room for experimentation with contested AI features in Europe is shrinking fast.

The court's ruling

The Amsterdam court has effectively drawn a line around one of the most toxic uses of generative AI: creating "undressed" images of real people without their permission. This is not about an artistic filter or stylization, but about content that can imitate intimate photographs and be used against a person's will. For xAI, this is not an abstract warning but a direct ban on generating and distributing such material in the Netherlands.

The fact of court intervention alone shows that the issue has moved beyond debates about product ethics.

It also matters that the ruling covers not only the publication of finished images but the very ability to create them. That changes the framing of the dispute: the court treats the feature as a potential tool for violating rights, not merely as content that someone may later post online. This approach raises pressure on the developers of models and interfaces. If a feature enables abuse from the outset, courts and regulators are increasingly ready to restrict it at the product level rather than wait for mass complaints from victims.

Why it matters

The story of "undressing" people with AI provokes a particularly strong reaction because the technology hits several sensitive areas at once: privacy, consent and reputation. Even if the image is entirely synthetic, the harm to the person can be very real, from harassment and blackmail to professional consequences. That is why courts view such tools not as neutral image generation but as a mechanism that makes it easier to produce non-consensual intimate content. For European legal culture, this is an especially sensitive topic.

At the product level, objections to such features usually come down not to a single detail but to a whole set of risks. The problem is not only that a contested result can be obtained, but how cheaply, quickly and at what scale this happens in practice through an ordinary interface. In the case of the Dutch ban, four questions are especially important, and any AI platform will inevitably face them:

• can an image of a real person be created without their consent;
• how easy is it to distribute such content afterward;
• who is responsible for the result: the user, the platform, or both;
• are there real technical barriers, rather than formal warnings in the interface.

What changes

For xAI, the ruling is a signal that contested model capabilities will now be judged not only by general safety promises but by the specific user scenario. If a tool can quickly produce harmful content, the claim of a "neutral technology" may no longer be enough. Companies will have to prove that restrictions are built into the product, work in practice and cannot be bypassed in a couple of prompts. Otherwise, local restrictions in one country may become a template for new lawsuits, investigations and bans in other European markets.

For the market as a whole, this is another example of generative AI ceasing to follow the "launch first, sort it out later" logic. The closer a model gets to the body, identity and personal data, the higher the likelihood of a direct legal response. This is especially true in Europe, where the protection of dignity and privacy is often placed above product speed. This applies not only to xAI: any developer adding features on the edge of sexualized, fake or degrading content now has a clearer marker of where the red line runs.

What it means

The ban against xAI shows that the fight over AI is now taking place not only at the level of big laws but at the level of individual product features. If a tool makes unauthorized humiliation or exploitation of people too easy, a court can intervene in a targeted and swift way. For the industry, this is bad news for murky growth mechanics and good news for users, whose rights have often lagged far behind the technology.