This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
A deepfake ecosystem on Telegram utilized the DeepNude AI software to allow users to generate non-consensual nude imagery of women and minors from submitted photos. The service, which operated via bots, gamified the harassment by allowing users to trade and rate images. Despite the original DeepNude software being taken down by its creators, the technology was reverse-engineered and deployed in this automated, paid-access format, impacting over 100,000 victims.
Seven channels were connected in a Telegram ecosystem centered around letting subscribers, as a paid service, generate non-consensual deepfake nudes using a bot from submitted photos of women, including underage girls and women who they know in real life.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Using AI systems to gain a personal advantage over others such as through cheating, fraud, scams, blackmail or targeted manipulation of beliefs or behavior. Examples include AI-facilitated plagiarism for research or education, impersonating a trusted or fake individual for illegitimate financial benefit, or creating humiliating or sexual imagery.
Also classified as
The Telegram bots generated and widely distributed sexually explicit, non-consensual, and potentially underage pornographic content in public channels.
The incident involves the targeted creation of humiliating sexual imagery of specific victims (women they know in real life) for personal gratification and harassment, which fits the boundary rule for synthetic content targeting specific victims.
Human
Due to a decision or action made by humans
Intentional
Due to an expected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The harm was caused by human actors who intentionally reverse-engineered, deployed, and used the DeepNude AI software to generate non-consensual pornography.
Added categories peak at 3: Substantial (Child Sexual Exploitation & Abuse (CSEA), Direct)
The deployment of automated Telegram bots utilizing reverse-engineered DeepNude software to generate non-consensual pornography affected over 100,000 women and minors. While representing a massive violation of privacy and human rights, the national security impact remains minor as it lacks state-actor involvement, kinetic threats, or critical infrastructure disruption.
The DeepNude AI system operates as a tool that performs a specific, well-defined function of generating a nude image from a user-submitted photo. It operates on direct input without context retention or autonomous decision-making.
Limited Risk: The system generates deepfakes, which are explicitly classified under the EU AI Act as posing a moderate risk requiring specific transparency obligations to ensure users are aware the content is AI-generated.
The primary function of the DeepNude AI system is to generate realistic deepfake nude images of women by digitally removing their clothes from submitted photos.
The reports from Sensity and investigative journalists provide clear, consistent, and detailed evidence of the Telegram bot's operation, the underlying DeepNude technology, the scale of the impact (100,000+ victims), and the demographic details of the users and victims. There is high certainty regarding the primary harms (harassment, non-consensual pornography, and privacy violations).
None
Technology
Reverse-Engineered GAN Software
DeepNude code was reverse-engineered and shared on open repositories.
Automated Bot Integration
Telegram bots automated image generation, making the tech highly accessible.
Rudimentary Video Undressing
Algorithms are expanding to process video frames sequentially.
Data Inputs
Unprotected Social Media Photos
Target images are easily harvested from public social media accounts.
Lack of Input Verification
No automated filters check if input photos belong to consenting individuals.
Human Factors
Malicious Intent of Users
Users seek to create nonconsensual porn for harassment or consumption.
Gamification of Harassment
Community rewards users with tokens for highly-liked nude creations.
Process and Methods
Automated Payment Systems
Bots process micro-payments automatically to remove watermarks.
Inadequate Content Moderation
Telegram failed to monitor and remove deepfake channels and bots.
Management
Platform Inaction
Telegram management ignored reports and requests to remove malicious bots.
VK Platform Vulnerabilities
Social media platforms struggled to proactively block deepfake communities.
Regulatory Environment
Absence of Legal Recourse
Lack of active prosecution or strong laws against AI revenge porn.
Inadequate Law Enforcement Action
Authorities failed to take tangible action against discovered networks.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
