Deepnudes are part of a larger problem: gender-based sexualized violence in the digital space

So far, there is no uniform definition of what exactly the term “digital (sexualized) violence” encompasses. Depending on the source, different terms are used to describe the same phenomenon: sometimes the term “gender-based violence in the digital space” is used, other times simply “digital violence,” which can also encompass acts that are not sexualizing or gender-related. At AlgorithmWatch, our understanding is based on the definition provided by the Federal Association of Rape Crisis Centres and Women’s Counselling Centres (bff): This definition classifies “digital violence” as an umbrella term for forms of gender-based violence that are committed using technological means or in the digital sphere. Digital sexualized violence occurs in the public digital space – for example, through the sharing of sexualizing content on social media platforms – as well as through the digitization of sexualized violence in private social relationships. This includes closed chats and messaging groups. Accordingly, “digital violence” cannot be viewed as entirely separate from “analog violence,” but rather is an extension of analog violence into the digital space. Among other things, it encompasses the following forms:

• Sexualizing deepfakes created and/or distributed without consent
• Image-based sexualized violence (for example, sending unsolicited sexual images such as “dick pics”), including the non-consensual sharing of intimate images (“Image-based Sexual Abuse,” or IBSA for short),
• Sextortion (blackmail with the use of intimate content),
• Sexualized insults, hate speech, and threats on social media and via messaging services,
• Cybergrooming (the targeted initiation of sexual contact with minors online),
• Stalking and doxxing with a sexualized component. Stalking refers to persistent, unwanted pursuit and threats that can escalate to psychological and physical violence. Doxxing refers to the publication of contact information or the creation of fake profiles, sometimes in connection with calls for violence.

Migrant women, LGBTQI+ people, and women in the public eye are disproportionately affected – in Germany and worldwide

In research and academia, non-consensual sexualized deepfakes are regarded as a form of gender-based violence embedded in patriarchal power and gender relations. Women and children are particularly affected. An intersectional perspective also shows that marginalized groups are disproportionately targeted: women with a migration history, girls, children, and adolescents, as well as queer and trans people, experience digital sexualized violence more frequently and in more severe forms. Women in the public eye (e.g., actresses, singers, female politicians, activists) are often the primary targets of non-consensual sexualized deepfakes, as their images are typically accessible for free via the internet. In some cases, the interplay of various forms of discrimination can exacerbate the violence, for example misogyny and racism. Such deepfakes must therefore also be recognized as a threat to democracy that targets specific groups and can only be addressed through a society-wide confrontation.

In Germany, the availability of reliable data on gender-specific digital violence is limited. According to police statistics, more than 90% of the victims of sexual offenses recorded in this country are female. The LeSuBiA study on unreported cases (2026) specifically examined digital violence for the first time – ranging from image manipulation and defamatory fake profiles to digital sexual harassment and stalking. It shows that in the past five years, nearly one in five women (18.3%) has experienced such violence “in the broader sense,” compared to 11.7% of men, with women with a migration history being disproportionately affected (22.2%). However, sexualized deepfakes were not recorded as a separate phenomenon in this study either.

A joint study by the Technical University of Munich and the organization HateAid (2025) shows that politically active women are disproportionately affected by digital violence. 63% of the women surveyed have reported experiences of digital violence, compared to 53% of men. It is particularly alarming that nearly two-thirds of the women affected have experienced gender-based violence in the form of sexism, misogyny, and misogynistic statements, while close to a quarter of them have received threats of rape. A survey by the Society for Civil Rights (GFF) among the German-speaking population aged 18 and older from 2021 showed how deeply digital violence affects young women in particular and how it shapes them in the long term: One in four young women has personally experienced digital violence; one in three fears that intimate images of her could be published on the internet without her consent.

A joint investigation by the tech magazine WIRED and the platform Indicator has systematically assessed the extent of sexualized deepfakes being spread in schools worldwide: Since 2023, at least 28 countries, among them Germany, report cases in which predominantly male adolescents at secondary schools attacked their female classmates with AI-generated nude images. A total of around 90 schools and more than 600 female students are affected, though the actual number of cases is likely higher. Some teachers have also been affected or have become perpetrators, as one case in the German state of Hesse shows. According to a UNICEF estimate, around 1.2 million children worldwide discovered sexualized deepfakes of themselves based on a study from 2025. Since the individuals depicted are minors, the content is legally classified as Child Sexual Abuse Material (CSAM).

The number of sexualized deepfakes is rising significantly worldwide. A 2023 study recorded a 550% increase in deepfake videos compared to 2019. 98% of these were sexual in nature, and 99% depicted women. A clear geographical and professional trend is also evident: those who become victims of deepfakes are predominantly active in the entertainment industry, while South Korea tops the list of countries with for 53% of all recorded sexualized deepfakes that are distributed online. The study attributes this to cultural stereotypes and geopolitical factors as well as specific target audiences. Furthermore, seven of the ten largest porn platforms were registered in 2023 to host sexualized deepfakes, which collectively garnered over 300 billion views. One in three deepfake tools that were examined allowed users to create sexualized content for free in less than half an hour.

Just how much this trend is being pushed through integration of generative AI into platforms and services can be seen in a study on the AI chatbot Grok, implemented on X (formerly Twitter): Between 5th and 6th January 2026, approximately 85% of deepfakes that were generated per hour – more than 7,800 images in total – were sexual in nature. Behind these deepfakes lies a growing industry that deliberately attempts to evade any responsibility. Providers of deepfake services hide behind foreign corporate headquarters and anonymous structures, while their services can be accessed effortlessly by using ordinary payment methods. The business model remains attractive – until economic pressure is applied: in 2020, the withdrawal of credit card support forced Pornhub to change its practices.

Screenshots from the "NST Detection Tool" developed by AlgorithmWatch, with which suspected nudify apps can be found in app stores. The images do not necessarily depict apps that can create sexualized deepfakes. Further tests are required to determine which features the apps provide.

What are the consequences for the victims and society?

The consequences of sexualized violence in the digital spere for those affected are severe. They suffer from significant psychological and physical stress, which can range from trauma – such as anxiety disorders, shame, social isolation, and suicidal thoughts – to health crises. In the case of non-consensual sexualized deepfakes, their widespread availability also adds the risk of normalizing sexualized violence against women. A cross-national study from 2022 shows that the emotional, psychological, professional, and social consequences (e.g., the threat to one’s economic livelihood due to the loss of collaborations and job opportunities) often continue to exist long after the original abuse. For those affected, this can lead to persistent anxiety about who has seen the images or who might see them in the future. According to another study, blaming victims and downplaying digital sexualized violence are particularly common among male, heterosexual respondents: Women are pressured into sexting (sending and receiving intimate images and videos of themselves), but are simultaneously held responsible for its consequences. The assumption that physical violence is more serious than “online” violence is also widespread – sometimes even among law enforcement agencies and first-response agencies, which contributes to such cases being taken less seriously. However, digital gender-based violence can often lead to offline assaults such as stalking or sexual harassment.

Deepfakes are also specifically used to intimidate women (and LGBTQI+ individuals) and push them out of public and political debates. This creates a “silencing effect”: the restriction of public expression, withdrawal from social media, or resignation from political office. Particularly women who are active in the public or in politics limit their participation or withdraw themselves entirely as a result of experiencing digital sexualized violence. A study from 2025 shows that in Germany, a significant proportion of female politicians affected by digital violence consider giving up public office or not taking up an office at all. This not only weakens the diversity of opinion in democratic discourse but also jeopardizes the political representation of women.

But this type of exclusion does not only affect the political sphere. According to a report by WIRED, particularly in cases involving children and adolescents, the initial institutional response determines whether they are willing to confide in adults in the future. Such experiences in early life have a lasting influence on trust in public institutions, in the rule of law, and in one’s own self-efficacy.

All of this harm is exacerbated by gender-specific double standards, as sexualized content stigmatizes women and LGBTQI+ individuals more severely than men – which also influences who is even considered “affected.” Sex workers are often overlooked in this context, even where their work material is used without consent – frequently through scraping it from porn platforms – to create non-consensual deepfakes, thereby depriving them of control over their own image, their source of income, and their ability to defend themselves. Among them, this lack of protection affects marginalized women such as asylum seekers the hardest, as they can risk losing their residency status, for example, in places where sex work is prohibited.

Further studies show that the often humorous framing of deepfake content on social media shifts responsibility away from platforms and towards individuals (Gosse & Burkell, 2020; Cochran & Napshin, 2021). The media also tends to portray digital violence through gender-specific deepfakes as less relevant compared to deepfakes used for disinformation. An experimental study found that sexualized deepfakes are judged more leniently when they involve a well-known public figure or when the content was created solely for private sexual gratification rather than for public distribution.

The structural dimensions of motivations and male dominance in online spaces

Research into the specific process of creating and distributing non-consensual sexualized deepfakes is still in its early stages, but has in recent years identified recurring patterns – both in terms of the motivations of abusers and the perception within society. Simultaneously, many studies emphasize that individual motivation alone is not a sufficient explanation for the dissemination of such content. As of now, research into what motivates consumers to engage with deepfakes is limited as well. Only one aspect emerges time and time again: the perpetrators are predominantly men.

The motives of both adolescent and adult abusers are wide-ranging, from sexual desire or curiosity via testing one’s courage to acting out revenge. It is becoming increasingly clear that sexual gratification is not necessarily the primary goal. Instead, exercising power is taking center stage, including humiliation, degradation, and social control, reputation damage, as well as revenge (Vasist & Krishnan, 2023).

The motivation for sexualized violence in the digital sphere is part of a broader gender-related context: The internet continues to be shaped by sexist and misogynistic patterns, while misogynistic radicalization is advancing – particularly on social media platforms. This is evident, for example, in the so-called “manosphere”, a loose network of online communities that cultivates a rigid understanding of masculinity. Within this echo chamber, the notion is propagated that progress in gender equality and women’s rights comes at the expense of men. Studies of relevant online communities paint a similar picture: research in forums on GitHub, Reddit, and comparable platforms describe the creation of sexualized deepfakes as embedded in a culture of “toxic masculinity”, in which technical expertise is used to boast male dominance (Newton & Stanfill, 2020; Winter & Salter, 2020; Kalpokas et al., 2024). Recurring patterns in these communities include reducing the depicted person to mere visual material, denying the non-consensual sexualized nature of the content, and displaying a pronounced indifference to the harm that those affected have to suffer. Deepfake violence appears less as an isolated act of aggression and more as a collectively staged, hyper-masculine assertion of entitlement over the female body, representing a “new form of digital sexual abuse.”

How do laws protect against “deepnudes,” and what role do AI providers and platforms play?

The current legal landscape consists of a combination of European and national regulations. Many legal building blocks already exist or are in the works. However, only their consistent interplay, an effective enforcement, the closing of existing loopholes, and a societal change in mindset can effectively protect those affected.

Platforms, AI providers, and perpetrators are the central actors addressed in legislation when it comes to non-consensual sexualized deepfakes. The former play a key role in the dissemination and are regulated by the European Digital Services Act. Operators of AI systems are the primary focus of the law, in particular with regard to the technological production of sexualized content, which is to be regulated by expanding responsibilities under the European AI Act. Perpetrators, in turn, are the main reason for which the German government is currently developing the Law Against Digital Violence, which, according to the current draft, criminalizes both the creation and the distribution of non-consensual sexualized deepfakes.

The Digital Services Act (DSA) requires very large online platforms (those with more than 45 million monthly users in the EU, also known as VLOPs) to proactively assess and mitigate so-called “systemic risks”. This explicitly includes gender-based violence as well as violations of human dignity and privacy. In member states where non-consensual sexualized content is classified as illegal, the DSA obligations to address and report such content apply to all online services, not just the major platforms.

Article 5 of the AI Act contains a list of practices for which the use of AI systems is prohibited in the EU. During a review process of the regulation in 2026 (a so-called Omnibus procedure), it was decided to expand the scope to include an explicit ban on AI tools that, without safeguards, enable the creation of non-consensual intimate content.

Under international law, the Council of Europe’s Istanbul Convention, which has been binding in Germany since 2018, mandates signatory states to protect against and combat violence against women in all forms. Furthermore, EU member states are obliged under the EU Directive on Combating Violence against Women and Domestic Violence (2024) to criminalize gender-based digital violence and to consider gender-based motives as grounds for imposing harsher penalties. This directive must be transposed into national law by member states by June 2027.

At the national level, the Federal Ministry of Justice and Consumer Protection (BMJV) recently presented the draft for a Digital Violence Protection Act. So far, German law does not include a specific criminal offense for non-consensual sexualized deepfakes – even though the federal government at the time issued an initial policy paper already in 2023. The Law Against Digital Violence announced by the Federal Ministry of Justice has been under discussion in the Federal Cabinet since the end of May 2026.

Victims in Germany also have a right to legal or psychological counseling, which, although enshrined in the Federal Government’s Violence Assistance Act, will not take effect before 2032. Until then, women in Germany are left to handle the consequences on their own.

The implementation of existing laws and planned legislative initiatives also leaves room for improvement, as well as the willingness of responsible authorities for enforcing them. After all, an effective response from the justice system does not depend solely on defining appropriate criminal offenses. It is also a matter of whether the police, prosecutors, and courts take victims seriously, believe them, and have the necessary specialized knowledge regarding AI-enabled acts of violence. The gap between the formal legal framework and actual criminal prosecution is wide and ranges from victim-blaming (attributing responsibility to victims of a crime rather than to the perpetrators, also known as Reverse Victim and Offender), normalization and a lack of reporting channels, all the way to authorities themselves making it difficult or discouraging victims from filing reports – often under the pretext that the perpetrators cannot be identified. In this respect, the assault experienced by victims is followed by an institutional response that downplays the injustice rather than acknowledging it, thereby afflicting pain to the victims twice over. Surveys from Germany highlight this lack of support structures: One in three respondents does not trust any official body to effectively address digital violence. 78% considered the platforms’ protective measures to be insufficient, while the reporting rate for digital violence among affected women is only 2.4%. In addition to political measures, targeted training for law enforcement agencies is therefore equally important.

Forms of violence such as non-consensual sexualized deepfakes are far more than a phenomenon of technological abuse: They extend patriarchal power structures into the digital realm, hit marginalized groups the hardest, and push women away from public and political debates. Curbing this violence requires not only the interplay of existing laws and new ones, but also their consistent enforcement – as well as a cultural shift in society that no longer downplays digital violence.

Further links and literature

The Office of the United Nations High Commissioner for Human Rights. (2026). Technology Facilitated Gender-Based Violence. https://www.ohchr.org/sites/default/files/documents/issues/women/genderandequality/2025-tool-technology-facilitated-gbv.pdf

Zorloni, L., AI Forensics, Open Society Foundations, Luminate, Limelight Foundation, & UCM. (2026). Harassment As Infrastructure: How Telegram’s design enables technology-facilitated gender-based violence in Italy and Spain. https://aiforensics.org/uploads/NCII%20Telegram%20-%20Harassment%20As%20Infrastructure.pdf

Burgess, M. (2026). The deepfake nudes crisis in schools is much worse than you thought. WIRED. https://www.wired.com/story/deepfake-nudify-schools-global-crisis/

Deutscher Juristinenbund e.V. (djb). Rechtlicher Handlungsbedarf bei nicht-einvernehmlichen sexualisierenden Deepfakes (2026). https://www.djb.de/presse/stellungnahmen/detail/st26-09#

LeSuBiA - Lebenssituation, Sicherheit und Alltag. Eine geschlechterübergreifende Bevölkerungsbefragung zur Gewaltbetroffenheit in Deutschland. (2026). https://www.bka.de/DE/UnsereAufgaben/Forschung/ForschungsprojekteUndErgebnisse/Dunkelfeldforschung/LeSuBiA/lesubia_node.html

HateAid, Koch, L., Voggenreiter, A., Steinert, J. I. (2025): Angegriffen & alleingelassen. Wie sich digitale Gewalt auf politisches Engagement auswirkt. Ein Lagebild. https://hateaid.org/wp-content/uploads/2025/01/hateaid-tum-studie-angegriffen-und-alleingelassen-2025.pdf

Bundeszentrale für politische Bildung: Feminizide und Gewalt gegen Frauen. https://www.bpb.de/themen/gender-diversitaet/femizide-und-gewalt-gegen-frauen/560023/digitale-gewalt-im-sozialen-nahraum

Flynn, A., Powell, A., Eaton, A., & Scott, A. J. (2025). Sexualized Deepfake Abuse: perpetrator and victim perspectives on the motivations and forms of Non-Consensually created and shared sexualized deepfake imagery. Journal of Interpersonal Violence, DOI: 10.1177/08862605251368834. https://doi.org/10.1177/08862605251368834

Lazard, L., Capdevila, R., Turley, E. L., Gilfoyle, K., & Stavropoulou, N. (2025). Deepfake Technology and Gender-Based Violence: A scoping review. Trauma Violence & Abuse, DOI: 10.1177/15248380251384271. https://doi.org/10.1177/15248380251384271

HateAid. (2025). Safe to engage: Securing democratic voices online — HateAid. https://hateaid.org/hass-hetze-politisch-engagierte-safe-to-engage/

Kalpokas, Ignas, et al. (2024). „Technology as a Threat or a Solution? The Challenges of Responding to Synthetic Media.“ Baltic Journal of Law & Politics, vol. 16, no. 2, Faculty of Political Science and Diplomacy and the Faculty of Law of Vytautas Magnus University (Lithuania), pp. 1-22. https://reference-global.com/article/10.2478/bjlp-2023-0010?tab=preview

GFF – Gesellschaft Für Freiheitsrechte e.V. Die Marie-Munk-Initiative. (2023). https://freiheitsrechte.org/themen/starke-grundrechte-fuer-eine-lebendige-demokratie/marie-munk-initiative

Flynn, A., Cama, E., Powell, A., & Scott, A. J. (2022). Victim-blaming and image-based sexual abuse. Journal of Criminology, 56(1), 7–25. DOI: 10.1177/26338076221135327 https://doi.org/10.1177/26338076221135327