This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
A sophisticated AI-generated deepfake video impersonating a prominent diabetes expert and a journalist has been circulating on social media in New Zealand. The scam promotes a fraudulent diabetes cure, causing victims to lose money and, more critically, encouraging them to discontinue medically necessary treatments. Despite reports to authorities, the scam persists as new variants of the video continue to emerge online.
A purported AI-generated deepfake video reportedly impersonated Auckland University diabetes expert Dr. Rinki Murphy, depicting her in a TVNZ interview with journalist Jack Tame promoting a fake diabetes cure. The alleged scam video circulated on social media in April–June 2025, prompting concern that New Zealanders were deceived into stopping prescribed medications. Despite reporting the incident, new variants of the video reportedly continued to emerge.
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 deepfake video exposed users to highly unsafe medical advice, encouraging diabetes patients to stop taking their prescribed medications.
Vulnerable patients trusted and relied on the deepfake video's medical claims, leading them to make unsafe health decisions.
The incident involves malicious actors using AI-generated deepfake videos of trusted public figures to run a targeted financial scam and manipulate vulnerable diabetes patients.
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 scam was intentionally created and deployed by human malicious actors using AI deepfake technology post-deployment to defraud individuals.
Added categories peak at 3: Substantial (Epistemic Harm, Indirect)
An AI-generated deepfake scam in New Zealand impersonated a prominent journalist and medical expert to promote a fraudulent diabetes cure. While posing a direct health risk to individuals who stopped medication and causing minor financial losses, the incident represents a minor national security threat primarily concerning the erosion of societal trust in media and health institutions.
The reports only mention that the videos were AI-generated and that the AI technology is becoming more sophisticated, but they do not describe the specific operational mode or characteristics of the AI system used.
Risk Level 3: The incident involves AI-generated content such as deepfakes, which are classified as Limited Risk under the EU AI Act, requiring transparency obligations so users know they are interacting with synthetic media.
The AI system was used to generate a highly realistic deepfake video and voice clone of a doctor and a journalist to deceive viewers.
The reports provide consistent accounts of the deepfake scam, identifying specific victims, financial losses, and the nature of the deceptive content. The role of AI in generating the video is clear, and the primary domain classification of fraud/scams is highly unambiguous.
None. The incident is clearly a malicious deepfake scam targeting vulnerable individuals for financial gain.
Technology
Sophisticated Deepfake Generation
AI tools allow highly realistic cloning of voices and facial movements.
Inadequate Algorithmic Detection
Platform algorithms fail to automatically flag and block deepfake scam ads.
Data Inputs
Exploitation of Public Media
Scammers use public broadcast footage of journalists and medical experts.
Lack of Media Provenance Data
No digital watermarks or metadata to verify the authenticity of the video.
Human Factors
Trust in Public Figures
Victims trust recognizable TV presenters and medical professionals blindly.
Desire for Chronic Illness Cure
Diabetes patients are highly motivated to find easy cures for their condition.
Lack of Digital Literacy
Vulnerable consumers fail to notice poor lip-syncing and red flags.
Process and Methods
Reactive Moderation Process
Platforms rely on manual reporting instead of proactive deepfake prevention.
Whack-A-Mole Takedown Loop
Deleted scam videos are immediately re-uploaded under different accounts.
Management
Prioritizing Ad Revenue
Social media platforms prioritize ad sales over strict advertiser verification.
Slow Incident Response
Ineffective escalation paths for victims and experts reporting identity theft.
Regulatory Environment
Weak Social Media Accountability
Lack of strict regulations holding platforms liable for hosting scam ads.
Inadequate Consumer Protection
Current laws fail to deter cross-border AI-generated financial scams.
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.
