For a long time, much of the conversation about AI has revolved around promises of productivity, creativity and automation. But there are uses that exhibit a much harsher reality: the possibility of generating non-consensual sexual content. It is not a minor detail, because it directly affects the privacy and dignity of the victims. The European Union has decided to respond to that specific point with a ban that already has political agreement.
The novelty. We are facing a ban that is integrated into a broader negotiation on the AI Act. The European Parliament explains that its negotiators and those of the Council have reached a provisional agreement to adjust several obligations of the rule, arguing to make it easier to comply without altering its risk-based approach. Within that package appears the piece that changes the game for these apps: the co-legislators have agreed to prohibit systems capable of creating this type of content. The final process, however, still requires formal adoption.
The role of Spain. The Spanish Government claims a direct role in the measure. Moncloa maintains that the ban agreed in Brussels comes from a proposal that Spain put on the table in January, after the controversy generated by the nudes of women and minors created with GrokX’s virtual assistant. According to that official version, the initiative achieved the support of the Union in mid-March to be included in the reform of the European Artificial Intelligence Law. Pedro Sánchez has also celebrated the agreement with a direct political message: “No more doing business by violating the dignity of our sons and daughters.”
The red line. The text identifies two especially sensitive categories. On the one hand, the systems used to create child sexual abuse material. On the other hand, those that allow the intimate parts of an identifiable person to be represented or shown in sexually explicit activities without consent. The ban, it should be noted, covers images, video and audio. Furthermore, the agreement is not limited to systems designed to generate this material: it also covers those who put them on the European market without reasonable measures to prevent it and those responsible for their deployment who use them for that purpose.
The key date. The ban already has a calendar on the table, and that changes the reading of the matter quite a bit. The official note sets December 2, 2026 as the deadline for companies to adapt their systems to the new rule. Before getting there there is an important formal step: the provisional agreement must be adopted by the European Parliament and the Council to enter European legislation. Even so, the path is marked, because they aim to close this procedure before August 2, 2026.
The fines. In the official statement we have not found reference to sanctions. However, the Irish law firm Matheson pointed out in April that both the Council and Parliament defended incorporating this practice into article 5 of the AI Act, the section reserved for prohibited practices. This requirement is relevant because, in the AI Act, article 99 provides for non-compliance with these prohibitions with administrative fines of up to 35 million euros or up to 7% of the company’s annual global turnover, if that amount is higher. For SMEs, a more limited criterion is contemplated.
Images | Xataka with Nano Banana
In Xataka | How to request to delete photos in which you have been naked sexualized with artificial intelligence
GIPHY App Key not set. Please check settings