This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
Two Canadian investors lost a combined $2.3 million to a sophisticated cryptocurrency investment scam that utilized AI-generated deepfake videos of Elon Musk and the television program Dragon's Den. The scammers used these videos alongside fabricated investment dashboards to build trust and persuade victims to transfer large sums of money. The report highlights the growing threat of industrial-scale fraud operations that leverage AI to manipulate victims.
Two Canadian investors in Ontario and Prince Edward Island reportedly lost a combined $2.3 million after being deceived by purportedly AI-generated deepfake videos impersonating Elon Musk and the television program Dragon's Den. The reported videos and fabricated online dashboards allegedly persuaded the victims to transfer funds into a fraudulent cryptocurrency investment scheme that displayed fake profits and blocked withdrawals.
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 deepfakes of public figures to conduct highly targeted financial scams and fraud 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 harm was caused by human scammers intentionally deploying AI-generated deepfake videos to defraud victims.
Added categories peak at 2: Minor (Epistemic Harm, Direct)
AI-generated deepfakes of public figures were utilized by Southeast Asian cyber fraud networks to scam Canadian citizens out of 2.3 million dollars. While representing severe financial and psychological harm to individuals and highlighting links to human trafficking, the incident poses minor direct threats to national security.
The AI system was used as a tool by human scammers to generate synthetic video and audio content (deepfakes) to deceive victims, operating on direct inputs without autonomous planning or agency.
Limited Risk: The report describes the use of AI-generated content such as deepfakes, which falls under Risk Level 3 and is subject to specific transparency obligations under the Act.
The primary function of the AI system involved was to generate highly convincing deepfake videos of Elon Musk and Dragon's Den to deceive victims.
The reports provide clear, detailed, and consistent accounts of the two victims, the specific financial losses, the platforms used, and the broader context of industrial-scale cyber fraud. The role of AI deepfakes is explicitly defined.
None
Technology
Realistic Deepfake Generation
AI-generated videos of celebrities like Elon Musk deceived victims.
Accessible Deepfake Tools
Cheap and accessible tools enable widespread creation of convincing scams.
Data Inputs
Manipulated Media Inputs
Use of realistic celebrity likenesses and voices as input data for ads.
Human Factors
Trust in AI Celebrity Endorsements
Victims believed fake AI endorsements from trusted figures like Elon Musk.
Overreliance on Video Authenticity
Victims assumed highly realistic video content on social media was authentic.
Process and Methods
Flawed Deepfake Detection
Social media platforms lacked effective processes to detect AI-generated ads.
Management
Inadequate AI Risk Monitoring
Platform management failed to monitor risks of AI-enabled ad fraud.
Regulatory Environment
Lack of Deepfake Regulations
Absence of strict legal frameworks governing AI-generated media.
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.
