This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
A Facebook video used deepfake technology to impersonate Croatian physician Alemka Markotić and TV host Zoran Šprajc, falsely claiming the physician had been murdered. The video and an accompanying fraudulent website, designed to mimic the Croatian Ministry of Health, were used to scam users into purchasing a fake joint-pain product called Hondrosol. This incident is part of a recurring pattern of AI-enabled fraud targeting public figures in Croatia.
A Facebook video purported to show Croatian infectious-disease physician Alemka Markotić and TV host Zoran Šprajc discussing her supposed murder and endorsing a joint-pain product called Hondrosol. The clip and linked article were reported to use AI-generated or manipulated media and to mimic Croatian government branding, directing users toward a scam purchase page.
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 fabricated a highly alarming and false news report claiming that a prominent physician had been murdered.
The incident involves malicious actors using AI-generated deepfakes of trusted public figures to impersonate them and run a fraudulent scam to sell a fake medical product.
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 initiated by human scammers who intentionally used deployed AI deepfake generation tools to create a fraudulent video and scam the public.
Added categories peak at 2: Minor (Epistemic Harm, Direct)
Malicious actors used AI deepfake technology to impersonate Croatian public figures, falsely claiming a prominent physician was murdered to promote a fraudulent joint-pain product through a fake Ministry of Health website. The incident represents a minor national security concern, characterized as a consumer fraud operation with localized impacts on public trust and government impersonation.
The AI systems were used as tools by malicious actors to generate synthetic video, clone voices, and translate/generate fraudulent text. The AI did not act autonomously or exhibit goal-directed planning on its own; it executed specific generation tasks based on direct human inputs.
Risk Level 3: Limited Risk. The report describes AI-generated content such as deepfakes, which under the EU AI Act are subject to specific transparency obligations to ensure users are informed that they are interacting with AI-generated content.
The primary technology used in the scam was deepfake video generation to make it appear as though the physician and TV host were speaking. Voice generation was also used to clone their voices.
The report from Faktograf is highly detailed, providing specific dates, names of the impersonated individuals, descriptions of the video content, and analysis of the fraudulent website. The fact-checking organization verified that the video was synthetic and that the claims were entirely false.
None. The evidence clearly points to an AI-enabled financial scam.
While the specific AI models used are unknown, this does not prevent reliable classification of the incident's domain, causes, and harms.
Technology
Advanced AI Video Generation
AI generated highly realistic video with synchronized lip movements.
High-Quality Voice Cloning
AI cloned voices of public figures with realistic tone and pitch.
AI Text Translation Errors
AI translation tool generated awkward Croatian text with gender errors.
Data Inputs
Abundant Public Voice Data
Large amounts of public broadcasts of the victims enabled voice training.
Public Video Material Availability
Existing TV show footage was used to train or overlay deepfake visuals.
Human Factors
User Susceptibility to Deepfakes
Users struggle to notice subtle lip-sync and audio anomalies in videos.
Trust in Public Figures
Scammers exploit public trust in famous doctors and TV hosts.
Urgency-Driven Decision Making
Scammers create artificial scarcity to pressure victims into buying quickly.
Process and Methods
Slow Social Media Moderation
Facebook failed to automatically detect and remove the deepfake video quickly.
Ineffective Domain Filtering
Lack of proactive blocking for suspicious URLs mimicking gov websites.
Management
Inadequate Platform Safeguards
Social media management prioritized engagement over proactive AI scam detection.
Delayed Institutional Response
Public health and police agencies react slowly to ongoing online scams.
Regulatory Environment
Lack of Deepfake Regulations
Absence of strict, enforceable laws targeting AI-generated scam videos.
Weak Online Pharmacy Oversight
Regulatory gaps allow illegal online drug distribution to persist.
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.
