This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
Multiple high-profile celebrities, including Tom Hanks, Gayle King, and MrBeast, reported that their likenesses and voices were used without consent in AI-generated deepfake advertisements. These fraudulent videos were circulated on social media platforms like Instagram and TikTok to promote products such as dental plans, weight loss supplements, and iPhone giveaways. The incidents highlight the growing misuse of generative AI for deceptive marketing and scams, prompting public warnings from the affected individuals and raising broader concerns about the impact of deepfakes on consumer trust and digital safety.
Deepfake technology was used to generate video advertisements featuring celebrities. Notable examples include the likeness of Tom Hanks touting a dental plan and another one in which the likeness of Gayle King touts a weight loss product. In each case, the individuals whose likenesses and voices had been deepfaked had not consented to their images and voices being used for the commercials.
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 reports highlight how the unauthorized replication of actors' voices and likenesses fuels ongoing labor disputes and fears of human creatives being devalued or replaced by AI.
The incident involved the unauthorized use of celebrity likenesses and cloned voices in deepfake advertisements to deceive social media users for financial gain through scams.
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 incident was caused by human malicious actors who intentionally created and deployed AI deepfakes post-deployment to scam social media users.
Added categories peak at 4: Severe (Epistemic Harm, Inferred)
Unauthorized use of prominent celebrity likenesses and cloned voices in AI-generated deepfake advertisements on social media platforms for commercial scams. Although showcasing the rapid progression and accessibility of synthetic media tools, the incident has minor direct implications for national security, primarily posing consumer fraud and localized epistemic trust challenges.
The reports describe generative AI being used to clone voices and manipulate video frames to match fake speech. The AI acted as a tool executing specific generation functions based on direct inputs provided by the creators of the scams.
Risk Level 3: The incident involves AI-generated content (deepfakes), which falls under Limited Risk due to transparency obligations requiring users to be informed that the content is synthetic.
The AI systems were used to generate realistic deepfake videos and clone the voices of celebrities to create fraudulent advertisements.
The reports provide consistent, detailed accounts of the celebrity deepfake incidents, including direct statements from the affected individuals and platforms. The nature of the scams and the technology used are clearly described across multiple reputable sources.
None
Technology
Advanced Voice Cloning
AI clones voices from short audio to make fake endorsements.
Realistic Faceswap Tools
Sophisticated AI easily overlays faces onto existing video clips.
Generative AI Accessibility
Powerful tools are widely available to make convincing deepfakes.
Data Inputs
Abundant Public Video Footage
Ample public footage of celebrities is used to train AI models.
Lack of Source Authentication
No tracking of original video origins before manipulation occurs.
Human Factors
Erosion of Public Trust
Viewers struggle to distinguish real videos from fake, causing skepticism.
User Susceptibility to Scams
Average social media users often fall for realistic fake advertisements.
Process and Methods
Inadequate Platform Moderation
Social media platforms fail to proactively block deepfakes before they spread.
Delayed Takedown Procedures
Platforms are slow to respond even after celebrities flag fake ads.
Management
Reduction in Ethical AI Research
Tech companies cut ethical AI teams as part of broader cost-cutting.
Prioritizing Growth Over Safety
Platforms focus on engagement rather than robust safety enforcement.
Regulatory Environment
Absence of Clear AI Laws
Few concrete laws exist globally targeting unauthorized deepfakes.
Unclear Intellectual Property Rights
Legal frameworks for protecting faces and voices as IP are still developing.
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.
