Executive Summary

AI-generated sexualized deepfakes are spreading exponentially – from 500,000 pieces of content in 2023 to 8 million in 2025. 99 percent of those depicted are women. Swiss data protection expert Jutta Oberlin warns: authorities cannot sustainably delete content once it has been distributed. Swiss criminal law offers only fragmented protection; a specific criminal offense for sexualized deepfakes is missing. The prominent case of German TV host Collien Fernandes illustrates the scale of the problem.

People

Topics

  • Deepfake pornography and digital violence
  • Legal gaps in criminal law
  • Platform responsibility and regulation
  • Gender-specific online violence

Clarus Lead

The problem is no longer marginal: authorities are reaching their limits in combating sexualized deepfakes. While the EU is pursuing a ban and the Swiss National Council is supporting corresponding demands, Oberlin reveals the uncomfortable truth – technically and legally it is already "hopeless". The discrepancy between political activism and actual enforceability marks a new dimension of digital violence that puts both legislation and platforms under pressure.

Detailed Summary

The scale of distribution has risen dramatically. Oberlin points to a 260-fold increase in AI-generated child sexual abuse material between 2024 and 2025. For adults, the dark figure is even larger: many victims don't even know that such material of them exists. Five Swiss influencers publicly revealed in 2026 that fake nude images were being distributed in Telegram forums – the reactions showed a mass phenomenon.

Oberlin identifies deepfake pornography as a new form of power exertion over women. She explains the psychological background: young women today are more independent and reject traditional partnerships. For men who view women as objects, the technology offers a simple means of exerting power. This is not just a technical issue, but a social one.

The Swiss legal situation is inadequate. There is no specific criminal offense for sexualized deepfakes. Existing articles such as "identity fraud" were created for completely different cases (such as grandparent scams) and do not cover the sexual component. Oberlin calls for a qualification of the law with a higher sentencing framework. Child sexual abuse material deepfakes are prosecuted (ex officio offense), but for adult material, pornography is generally permitted – victims must file a complaint, which many do not do because they don't even know that material exists.

Platforms bear significant responsibility. They have created the technical capabilities without appropriate security mechanisms. Oberlin criticizes: companies invest in product development and growth, not in compliance and user protection. The EU is planning a reporting requirement; Switzerland does not have one. A technical limit (photos uploaded only with consent) would make sense but is currently difficult to implement.

Regarding enforcement of a ban, Oberlin remains skeptical: the technology is already widespread, and perpetrators will find ways – including on the darknet – to create deepfakes. Law lags behind technology by a multiple. Laws must be formulated in a technology-neutral manner (as with the EU AI Regulation), but this makes concrete regulations more difficult.

One aspect: deepfakes disproportionately affect women, not only because they are more frequently targeted, but because society associates shame with them. With male politicians (Obama, Trump), the material is often dismissed as funny. Oberlin sees a gender discourse here: laws can help create shame awareness among perpetrators – but only if victims use them and file complaints, which takes courage.

Core Statements

  • Mass phenomenon: 8 million sexualized deepfakes in 2025 (99% women); authorities cannot sustainably delete distributed content.
  • Legal Gap in Switzerland: No specific criminal offense; identity fraud statute does not cover sexual component; deepfakes of child sexual abuse material are prosecuted, adult material is not.
  • Platforms as Catalyst: Companies do not invest in compliance and user protection; reporting requirement (as exists in the US for child sexual abuse material) is missing in Switzerland.
  • Gender-Specific Violence: Deepfakes are a form of power exertion; women are subject to shame, men often are not – legal and cultural asymmetry.
  • Skeptical Enforcement: a ban alone does not help; technology is already widespread; laws must be technology-neutral and lag behind innovation.

Critical Questions

  1. Source Validity: Oberlin cites 8 million deepfakes in 2025 and a 260-fold increase in AI-generated child sexual abuse material 2024–2025 – do these figures come from independent organizations (Internet Watch Foundation), or are there other estimates that deviate significantly?

  2. Conflicts of Interest: Oberlin worked at Google for a long time and publishes academically. To what extent do her former or current connections to tech companies influence her assessment of platform responsibility and technical solutions?

  3. Causality – Gender: Oberlin argues that deepfakes are a form of power exertion over women because young women are more independent. Could the problem also be explained by other factors (anonymity, impunity, technical ease) independent of the gender aspect?

  4. Feasibility – Darknet: Oberlin explains that laws would be ineffective because perpetrators would migrate to the darknet. Does this contradict her own call for stricter laws? Under what conditions would she consider a ban effective?

  5. Platform Responsibility – Technical Feasibility: She criticizes that platforms do not implement a consent-based upload system. How realistic is this solution technically and from a data protection perspective, and who would perform consent verification?

  6. Victim Bias: The article is largely based on statements from Oberlin. How many affected persons were interviewed? Is the perspective of law enforcement authorities or tech companies represented?

  7. Comparability – Countries: Oberlin mentions EU plans and German cases, but rarely concrete regulations in other countries. Are there countries that already have functioning criminal offenses, and what can these teach us?

  8. Dark Figure: The article emphasizes that many victims don't know that material of them exists. How is this dark figure estimated, and could a platform reporting requirement make the problem appear larger (which might be politically undesirable)?

Bibliography

Primary Source: NZZ Gesellschaft – Interview by Jana Schmid and Elena Oberholzer, "Can Authorities Sustainably Delete Deepfakes Once Distributed? 'Hopeless', says a Data Protection Expert" (https://www.nzz.ch/gesellschaft/koennen-behoerden-einmal-verbreitete-deepfakes-nachhaltig-loeschen-chancenlos-sagt-eine-datenschutzexpertin-ld.10014534)

Supplementary Sources (referenced in text):

  1. Internet Watch Foundation – Report on AI-Generated Child Sexual Abuse Material 2024–2025
  2. Tamedia Publications – Experience Reports from Five Swiss Influencers (Spring 2026)
  3. Cantonal Police Zurich – Information on Victim Support and Content Removal
  4. EU Initiatives on Sexualized Deepfakes and Planned Bans
  5. Swiss National Council – Resolution on Deepfake Ban (2026)

Verification Status: ✓ 11.07.2026

This text was created with the assistance of an AI model. Editorial Responsibility: clarus.news | Fact-Check: 11.07.2026