This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
On March 16, 2022, a deepfake video of Ukrainian President Volodymyr Zelenskyy was circulated online, falsely showing him calling for his country to surrender to Russian forces. The video was also broadcast on the hacked news channel Ukraine 24. The incident was quickly debunked by the Ukrainian government and removed by major social media platforms, but it highlighted the potential for AI-generated disinformation to be used as a weapon in armed conflict.
A quickly-debunked deepfaked video of the Ukrainian President Volodymyr Zelenskyy was posted on various Ukrainian websites and social media platforms encouraging Ukrainians to surrender to Russian forces during the Russia-Ukraine war.
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
Experts noted that the incident pollutes the information ecosystem and contributes to the 'liar's dividend,' making it harder for the public to trust authentic media.
The deepfake was deployed as part of a coordinated disinformation and propaganda campaign aimed at manipulating public opinion and population-level behavior during a war.
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 (hackers and propagandists) who intentionally created and distributed the deepfake video to deceive the public during an armed conflict.
Added categories peak at 3: Substantial (Epistemic Harm, Inferred)
A deepfake video of Ukrainian President Zelenskyy falsely calling for surrender was broadcast via a hacked news channel and social media. While the low quality and rapid debunking limited its immediate impact, it represents a first-of-its-kind use of AI-generated media as an information warfare weapon during active armed conflict.
The AI system functioned as a Tool, specifically a deepfake generation program used to perform the well-defined function of mapping Zelenskyy's face and synthesized voice onto another body based on direct inputs.
Limited Risk: The report describes the creation and dissemination of a deepfake video, which is explicitly classified under the EU AI Act as a limited risk system requiring transparency obligations so users know they are interacting with AI-generated content.
The AI system was used to generate a synthetic video of Volodymyr Zelenskyy's face ('faceswap') and a synthetic version of his voice to make it appear as though he was delivering a surrender speech.
The reports provide consistent, detailed, and well-sourced accounts of the incident from multiple reputable news organizations. The facts regarding the creation, distribution, hacking of Ukraine 24, and subsequent debunking of the deepfake are clear and uncontradicted.
None. The event is universally recognized as an attempted political disinformation campaign using deepfake technology.
Technology
Accessible Deepfake Software
Faceswap and AI voice synthesis tools are readily available to adversaries.
Low-Resolution Masking
Low quality was used intentionally to hide rendering glitches in the video.
Stilted AI Voice Synthesis
AI voice cloning generated a robotic but passable audio track of the leader.
Data Inputs
Abundant Source Material
High volume of public video of the target was used to train the AI model.
Hacked Broadcast Feed
Hackers injected fake text into the news ticker to support the deepfake.
Human Factors
Cognitive Fog of War
High-stress conflict makes the public more susceptible to panic and lies.
The Liar's Dividend
The existence of fakes makes citizens doubt the authenticity of real media.
Process and Methods
Vulnerable News Infrastructure
Lack of robust security allowed hackers to breach the TV station's systems.
Reactive Content Moderation
Social media platforms removed the video only after it started spreading.
Management
State-Sponsored Disinformation
Adversaries integrated AI deepfakes into their broader military strategy.
Regulatory Environment
Inadequate Platform Mandates
No regulatory standards exist to force proactive deepfake identification.
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.
