This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
AI-generated deepfake videos were circulated on Facebook by pages 'Gazeta Bota' and 'Peja LIVE' ahead of the June 2026 Kosovo elections. These videos impersonated media outlets like RTK and 'Ekonomia Online' to present fabricated citizen interviews and political claims. The content successfully reached thousands of users, posing a risk to democratic norms and public trust in legitimate media.
Videos reported as AI-generated were allegedly circulated by Gazeta Bota and Peja LIVE on Facebook before Kosovo's June 7, 2026 elections, posing as citizen interviews and media reports. The clips reportedly manufactured voter sentiment by presenting synthetic interviewees as real citizens in Kosovo media-style election reports.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Using AI systems to conduct large-scale disinformation campaigns, malicious surveillance, or targeted and sophisticated automated censorship and propaganda, with the aim of manipulating political processes, public opinion, and behavior.
Also classified as
The use of fabricated media reports risks damaging trust in professional journalism and creating widespread informational confusion during an election campaign.
The incident involves the coordinated creation and distribution of AI-generated deepfakes to conduct a large-scale political influence campaign targeting the Kosovo electorate.
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 actors who intentionally created and distributed deepfake videos to manipulate public opinion ahead of the Kosovo elections.
Added categories peak at 3: Substantial (Epistemic Harm, Inferred)
AI-generated political deepfakes impersonating legitimate Kosovo media outlets were circulated on Facebook to manipulate public opinion ahead of the June 2026 elections. This incident highlights the growing threat of synthetic media to election integrity and public trust in democratic institutions, though it represents an established threat vector without physical impact.
The AI was used as a tool to generate synthetic audio and video (deepfakes) based on direct inputs from human creators, who then manually uploaded the finished videos to Facebook pages.
Risk Level 3: Limited Risk. The incident involves 'AI-generated content such as deepfakes' which, under the EU AI Act, are subject to transparency obligations to ensure users are informed that the content is synthetic.
The AI system's primary function was to generate realistic synthetic videos and matching synthetic audio of individuals to mimic real news broadcasts.
The report provides clear, specific details about three distinct occurrences of deepfake videos, including dates, publishing pages, specific claims made, and audience engagement metrics. The role of AI in generating these videos is explicitly verified by visual and audio analysis.
None. The intent to manipulate the election using synthetic media is highly clear.
Technology
Deepfake Generation Tools
AI tools allowed creation of highly realistic synthetic audio and video.
Synthetic Audio Synthesis
Voice cloning simulated real political statements convincingly.
Visual Rendering Flaws
Inaccurate lip-syncing and facial movements showed AI generation.
Data Inputs
Unauthorized Media Logos
Use of RTK and Ekonomia Online logos created false credibility.
Synthetic Backgrounds
Fake backgrounds like Pristina square simulated real locations.
Human Factors
Trust in Visual Logos
Audiences naturally trust professional journalism indicators like logos.
Lack of Source Verification
Users consumed and shared content without verifying the original source.
Unsupervised Consumption
Rapid spread occurred as users engaged without critical evaluation.
Process and Methods
Lack of Platform Labeling
Social media platforms failed to label deepfakes as AI-generated.
Anonymous Page Distribution
Pages like Gazeta Bota distributed content without transparency.
Management
No Editorial Accountability
Publishing pages lacked identifiable staff or ownership transparency.
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.
