This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
A high-profile Twitch streamer inadvertently exposed a website hosting non-consensual, AI-generated sexually explicit deepfakes of his colleagues during a live broadcast. This incident triggered widespread circulation of the imagery, causing significant psychological harm, harassment, and professional distress to the targeted female content creators. The event highlighted the lack of effective legal recourse for victims of AI-facilitated non-consensual intimate imagery.
Unauthorized, non-consensual deepfake pornography showing faces of high-profile female streamers and content creators was published on a subscription-based website, which gained notoriety after a male streamer was caught accessing the site.
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 incident involved the creation, distribution, and exposure of non-consensual sexually explicit synthetic media depicting real individuals.
According to boundary rule B1, when a public likeness is sexualized or a person is harassed/degraded with synthetic content, it is classified under 4.3 if it involves targeted humiliation of 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 primary risk and harm were caused by the intentional actions of humans: the creator who generated the non-consensual deepfake pornography and the user who purchased and broadcasted it.
Added categories peak at 2: Minor (Psychological Harm, Direct)
An incident involving the creation and viral spread of non-consensual, AI-generated sexually explicit deepfakes of prominent female streamers. While causing severe psychological and reputational harm to the individuals targeted, the incident represents a localized harassment and regulatory challenge rather than a threat to national security.
The AI systems used (face-swapping apps and machine learning software) operated as tools. They performed specific, well-defined functions of mapping one face onto another body based on direct user inputs, without exhibiting autonomous goal-directed behavior or independent planning.
Limited Risk: The incident involves AI-generated content (deepfakes). Under the EU AI Act, deepfakes are classified as limited risk and are subject to specific transparency obligations, requiring creators to disclose that the content is artificially generated or manipulated.
The AI technology was explicitly used to generate realistic synthetic videos and images by swapping the faces of female streamers onto adult performers' bodies.
The reports provide consistent, detailed, and cross-verified accounts of the incident from multiple reputable media outlets (NBC, Wired, CNN, Kotaku). The perspectives of the victims, the perpetrator, the platform, and legal experts are thoroughly documented, leaving no major ambiguities regarding what occurred and the harms caused.
None; the facts of the non-consensual deepfake creation and subsequent stream leak are undisputed.
Technology
Accessible AI Face-Swapping Tools
Easy-to-use AI tools allow anyone to generate realistic deepfakes quickly.
Generative Models Trained on Web Data
Models trained on internet images reflect and amplify societal sexualization.
Lack of Robust AI Output Filters
Open-source models like Stable Diffusion lack absolute output guardrails.
Data Inputs
Unsanctioned Public Image Scraping
Publicly available photos and streams are scraped to train deepfake models.
Explicit Training Data Biases
Billions of internet images used in training default to sexualizing women.
Human Factors
Viewer Curiosity and Demand
Users paying for non-consensual content fuels the deepfake creator market.
Careless Exposure on Stream
A streamer accidentally showed a deepfake subscription site to thousands.
Lack of Online Literacy in Families
Victims' family members struggle to distinguish deepfakes from reality.
Process and Methods
Difficult Content Removal Process
Takedowns are expensive, slow, and fail to stop screenshots from spreading.
Inadequate Platform Moderation
Platforms struggle to detect and moderate deepfake hashtags and links.
Management
Move Fast and Break Things Ethos
AI companies release powerful tools without considering safety risks to women.
Delayed Platform Policy Response
Twitch took over a month to update policies regarding synthetic NCEI.
Monetization of Exploitative Sites
Subscription-based sites profit directly by hosting non-consensual porn.
Regulatory Environment
Absence of Federal Deepfake Laws
US federal law lacks specific protections, leaving victims with no case.
Section 230 Safe Harbors
Section 230 protects site owners from liability if they act in good faith.
Fragmented State Legislation
Only a few states explicitly criminalize or allow civil suits for deepfakes.
AI system involved
An AI system was involved, but the reports do not identify it.
Identified by automated extraction; can differ from the alleged developers above.
