This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
Multiple reports describe a rising trend of romance and employment scams utilizing AI-generated deepfake videos of celebrities to gain the trust of victims. In one instance, a woman's mother was targeted with a fake job offer and a deepfake video as 'proof' of the scammer's identity. Other victims have suffered significant financial losses, including one individual who lost £700,000, after being manipulated by AI-generated content and impersonation tactics.
A Reddit user reported in May 2025 that her mother is being groomed by scammers using an AI-generated deepfake video impersonating actor Owen Wilson. The scam reportedly began on a game app (Yahtzee with Friends) and shifted to WhatsApp voice calls, with the victim promised a fake Warner Bros job paying small sums via Cash App. The scammer allegedly claims he will buy a house for the victim's family as part of the deception. The video is cited as the main “proof” of authenticity.
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.
The incident involves malicious actors using AI-generated deepfake videos of celebrities to conduct highly targeted romance and financial scams against 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 scams were intentionally executed by human malicious actors using deployed AI deepfake generation tools to defraud victims.
Added categories peak at 3: Substantial (Epistemic Harm, Inferred)
This incident involves the use of AI-generated celebrity deepfakes to conduct highly targeted romance and financial scams across multiple nations, resulting in significant individual financial losses and psychological distress. While representing a serious criminal evolution in synthetic media exploitation, it carries low national security implications as it does not target critical infrastructure, state sovereignty, or national economic systems.
The AI systems were used as tools by human scammers to generate realistic deepfake videos of celebrities. The reports describe the technology as 'video generation models' and 'deepfake clips' used to trick victims, indicating the AI performed a specific, well-defined function on direct input without autonomous planning.
Risk Level 3: Limited Risk. The reports describe the use of 'AI-generated videos' and 'deepfake clips' of celebrities, which falls under the category of AI-generated content requiring transparency obligations under the Act.
The primary function of the AI systems described was to generate highly realistic deepfake videos of celebrities to deceive victims.
The reports provide consistent, detailed accounts of multiple victims targeted by celebrity deepfake scams, including specific financial losses, platforms used, and psychological impacts. The role of AI as a tool for generating deepfake videos is clearly stated across all sources.
None. The events are clearly romance and financial scams leveraging synthetic media.
Technology
State-of-the-Art Video Gen
Models like Veo 3 generate highly realistic videos that deceive victims.
Deepfake Audio and Video Synthesis
AI synthesized robotic voice and static eyes that mimic real celebrities.
High-Fidelity Image Manipulation
Scammers use AI to doctor selfies and IDs to look highly convincing.
Data Inputs
Publicly Available Celebrity Data
Abundant online photos and videos of celebs are used to train and feed AI.
Social Engineering Information
Personal details like family names are harvested to tailor the scam.
Fake ID and Document Images
Doctored driving licenses and certificates used to build false trust.
Human Factors
Cognitive Vulnerabilities
Victims suffering from depression or isolation are highly susceptible.
Emotional Manipulation
Flattery and romance triggers exploit victims' emotional desires.
Lack of AI Literacy
Victims fail to recognize robotic tones, static eyes, and pixelation.
Process and Methods
Scaled Phishing via AI
AI allows scammers to automate and scale up highly convincing phishing.
Cross-Platform Scam Operations
Scammers shift victims from games to WhatsApp, CashApp, and voice calls.
Financial Siphoning Tactics
Using non-refundable gift cards and small CashApp payments to build trust.
Regulatory Environment
Lack of Deepfake Regulation
Insufficient legal frameworks to stop the creation and spread of deepfakes.
Cross-Border Enforcement Limits
Jurisdiction issues make tracking international scammers highly difficult.
Weak Platform Moderation Laws
Gaps in laws fail to hold platforms accountable for hosting fake accounts.
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.
