This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
The report details the widespread use of AI to create non-consensual pornographic deepfakes, specifically targeting women. The victim, Breeze [name withheld], experienced severe psychological trauma after her image was used in over 800 deepfake links. The article highlights the role of search engines and platforms in monetizing this content and the lack of clear legal recourse due to Section 230 protections.
Porn sites are alleged to have used AI-generated images of Breeze Liu without her consent, leading to severe emotional distress. Liu discovered a video of herself on Pornhub, which was then deepfaked and spread across over 800 links. Despite efforts to remove the content, many sites refused to comply, perpetuating the violation and exploitation of her image.
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 proliferation of non-consensual pornographic deepfakes online exposes users and victims to highly toxic, abusive, and sexually explicit content.
The report also describes a deepfake robocall of Joe Biden during the New Hampshire primary and a fake endorsement video of Taylor Swift, which represent coordinated attempts to influence public opinion and political processes.
The primary incident involves the creation of humiliating and sexual deepfake imagery targeting a specific victim for harassment and monetization, which fits the boundary rule for targeted exploitation.
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 malicious human actors intentionally using deployed AI generation tools to create and distribute non-consensual deepfakes of victims.
Added categories peak at 3: Substantial (Psychological Harm, Inferred)
The incident details the widespread use of generative AI to create non-consensual pornographic deepfakes and political disinformation, such as a fake Joe Biden robocall. While primarily a severe societal and civil rights issue, the technology also presents minor national security concerns regarding election integrity and information warfare.
The deepfake generation software operates as a tool, performing the specific function of generating synthetic images and videos based on direct inputs of real people's faces without independent planning or context retention.
Risk Level 3: Limited Risk. The report describes the generation of deepfakes, which are explicitly classified under Limited Risk, requiring transparency obligations to ensure users are aware the content is AI-generated.
The primary function of the AI systems described is the generation of deepfake videos and images of individuals without their consent.
The report provides clear, first-hand testimony from a victim of deepfake abuse, along with statistical context about the scale of the issue and specific examples of political deepfakes.
None
Technology
Rapid Deepfake Tech Advancement
AI-generated imagery technology has advanced much faster than legal frameworks.
Identity Manipulation Tools
AI tools enable easy manipulation of real faces to generate realistic porn.
Data Inputs
Non-Consensual Image Harvesting
Real faces and identities are harvested and used without consent for deepfakes.
Human Factors
Shame Silencing Victims
Victims are humiliated and reluctant to speak out, perpetuating the problem.
Process and Methods
Ineffective Content Takedowns
Malicious platforms refuse to respond or take down non-consensual videos.
Search Engine Indexing
Search engines direct traffic to deepfake sites and profit from their ads.
Management
Prioritizing Profit Over Safety
Search engines accept ads from deepfake companies to generate profit.
Corporate Indifference
Tech executives ignore the social harms of indexing deepfake pornography.
Regulatory Environment
Absence of Federal Laws
There is no federal law that clearly covers non-consensual AI deepfakes.
Section 230 Liability Shield
Section 230 protects tech platforms from civil lawsuits for hosting content.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
