The European Union has also taken decisive action. EU member states and European Parliament members reached a political agreement to revise the Artificial Intelligence Act, explicitly prohibiting AI systems from generating non-consensual sexual deepfake content and child sexual exploitation material. This is the first EU legislation to explicitly ban "nudifier applications". The revised provisions require mandatory watermarking of all AI content, and the prohibition covers any image, video, or audio exposing intimate parts of a real person without their consent, exempting only purely fictional AI-generated characters. The immediate catalyst for this revision was the Grok scandal, with analysis showing that during a 24-hour period in early January, Grok was generating over 6,700 sexually suggestive images per hour, equivalent to 84 times the total output of the world's five largest deepfake websites combined.
The primary technology driving these applications is the . A GAN consists of two neural networks—a generator and a discriminator—that work in tandem. The generator creates synthetic images, while the discriminator critiques them for realism. Through this adversarial back-and-forth, the generator learns to produce highly convincing, forged imagery of a person without their clothing. These models are trained on vast datasets of both clothed and unclothed human figures, learning to realistically synthesize and replace a person's attire with simulated skin. Undress AI
While Undress AI has raised concerns about its potential misuse, it also has several legitimate applications: The European Union has also taken decisive action
The Undress AI phenomenon is closely tied to the emergence of Explainable AI (XAI), a field focused on making AI systems more transparent, accountable, and interpretable. XAI aims to provide a deeper understanding of AI decision-making processes, enabling humans to comprehend and trust AI outputs. Techniques like model-agnostic interpretability, attention mechanisms, and model explainability have become increasingly popular, allowing researchers to "undress" AI systems and reveal their inner workings. The revised provisions require mandatory watermarking of all