This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
During the December 2025 Kosovo parliamentary elections, various social media accounts and pages used AI-generated content to spread misinformation. This included deepfake videos and images depicting fabricated political events, fake polls, and manipulated photos of public figures. The content was designed to influence voter perception and mobilize support for specific political entities, significantly impacting the political atmosphere.
Fourteen purportedly AI-generated TikTok videos were reportedly posted by NoFilterNews from December 6–8, 2025, using television-news styling to promote election and diaspora-mobilization claims about Kosovo's December 28 elections. The videos purportedly favored Vetëvendosje and reportedly reached hundreds of thousands of viewers, and is alleged to have appeared to distort audience perceptions of political support.
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 AI systems generated highly realistic false images and videos that misled social media users.
The incident involved coordinated campaigns using AI-generated deepfakes and images to conduct large-scale political disinformation aimed at manipulating public opinion and the parliamentary election in Kosovo.
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 intentionally using deployed AI generation tools to create and spread political disinformation during an election campaign.
Added categories peak at 3: Substantial (Epistemic Harm, Inferred)
During the December 2025 Kosovo parliamentary elections, unidentified actors utilized advanced generative AI tools, including OpenAI's Sora, to execute a coordinated disinformation campaign. By spreading highly realistic deepfakes of political figures and fabricated news broadcasts, the operation aimed to manipulate voter perception and undermine democratic processes, illustrating the growing threat of AI-enabled information warfare to national sovereignty.
The AI systems (Sora and image generators) operated as tools, performing the specific function of generating videos and images based on direct user prompts without independent planning or autonomous decision-making.
Limited Risk: The report describes the use of AI-generated content such as deepfakes, which under the EU AI Act are subject to transparency obligations to ensure users are aware they are interacting with AI-generated media.
The AI systems were primarily used to generate deepfake videos (using Sora) and deepfake images of politicians and events to spread political misinformation.
The reports from hibrid.info provide detailed, fact-checked examples of specific AI-generated images and videos, including the platforms used, view counts, and the specific political context of the Kosovo elections.
None. The intent to manipulate the election using AI-generated misinformation is clear and well-documented.
Technology
AI Video Generation Tools
Tools like Sora allowed easy creation of realistic TV-style news videos.
Deepfake Image Generators
Image models created realistic fake photos of political figures.
Imitation of Media Logos
AI generated characters holding microphones with distorted real media logos.
Data Inputs
Lack of Provenance Metadata
Digital media lacked verifiable origin data to prove authenticity.
Use of Real Likenesses
AI models utilized real political figures' faces to generate deepfakes.
Human Factors
User Confirmation Bias
Users accepted fake content because it aligned with their political views.
Ignoring AI Labels
Users ignored TikTok's 'AI-Generated' labels and took videos as real.
Lack of Digital Literacy
Audiences failed to notice visual distortions in AI-generated media.
Process and Methods
Inadequate Platform Moderation
Social platforms failed to proactively block deceptive political deepfakes.
Delayed Fact-Checking
Fact-checkers verified content only after it gained high view counts.
Impersonation of News Formats
Creators used professional TV formats to bypass typical user skepticism.
Management
Prioritizing Engagement
Platforms prioritized engagement, amplifying deceptive AI videos.
Inadequate Risk Assessment
Platforms failed to prepare for targeted AI disinformation campaigns.
Regulatory Environment
Lack of Election AI Regulations
No legal framework existed to ban deceptive AI media during elections.
Weak Platform Accountability
Social media platforms faced no legal penalties for hosting deepfakes.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
