This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
Hungarian opposition leader Magyar announced a criminal complaint against a political aide for publishing an unlabeled, AI-generated deepfake video on Facebook. The video falsely depicted Magyar claiming he would cut and tax pensions, an act intended to influence the upcoming 2026 parliamentary election. The incident highlights the risks of synthetic media in political discourse and the lack of mandatory disclosure labels for AI-generated content in the current regulatory environment.
In Hungary, opposition leader Péter Magyar alleged that Viktor Orbán aide Balázs Orbán published an unlabeled deepfake video on Facebook reportedly depicting Magyar as saying he would cut and tax pensions ahead of the 2026 parliamentary election. Magyar reportedly said he would file a criminal complaint and described the video as defamatory.
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 system was used to generate false and misleading video content depicting statements the politician never made.
The incident involves using an AI-generated deepfake to conduct targeted political disinformation aimed at manipulating public opinion and the upcoming parliamentary election.
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 a human decision to intentionally create and publish a defamatory deepfake video of a political opponent after the AI model was deployed.
Added categories peak at 2: Minor (Epistemic Harm, Direct)
A high-ranking Hungarian political aide published an unlabeled, AI-generated deepfake video of opposition leader Peter Magyar to influence the upcoming 2026 parliamentary elections. This incident highlights the growing threat of synthetic media in domestic information warfare and its potential to undermine democratic processes and exacerbate political polarization.
The AI system was used as a tool to generate a synthetic video (deepfake) of Peter Magyar. It performed a specific, well-defined function of video generation based on direct inputs without exhibiting autonomous planning or decision-making.
Risk Level 3. Limited Risk: The report describes the creation of an AI-generated deepfake video. Under the EU AI Act, deepfakes and AI-generated content pose a moderate risk and require specific transparency obligations, such as labeling.
The AI system was used to generate a 38-second synthetic video and audio of Peter Magyar to make it appear as though he made statements he did not.
The report clearly details the political context, the specific video published, the platform used, and the legal actions taken. The role of AI as a deepfake generator is explicit, and the political motivations are clear.
None
Technology
AI Deepfake Generation
AI tools enabled creation of highly realistic fake video of politician.
Lack of Platform Detection
Facebook lacked automatic mechanisms to detect and flag deepfake video.
Data Inputs
Target Audio and Video Data
Source media of Peter Magyar was used to generate the deepfake.
Human Factors
Malicious Political Intent
Political aide intentionally shared fake video to discredit an opponent.
Process and Methods
No AI Disclosure Labeling
The video was published directly to Facebook without any AI-generated tag.
Regulatory Environment
Delayed EU AI Act Mandates
EU AI Act disclosure rules are not active until 2026, limiting compliance.
Untested Court Precedents
Civil and criminal laws exist but court rulings on deepfakes are untested.
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.
