This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
A political promotional video in Taiwan used AI voice cloning to mimic President William Lai's voice criticizing the government's handling of a cooking oil scandal, sparking police investigations and intense political debate over deepfake regulation and free speech.
A political video titled "It's Out of Your Control" reportedly used an AI-generated voice resembling Taiwan President Lai Ching-te while criticizing the government's response to a cooking-oil safety scandal. Police said the voice could be mistaken for Lai and later confirmed it was AI-generated. Opposition figures defended the video as political expression while prosecutors opened a forgery investigation.
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 cloned voice generated false statements attributed to the president, creating misleading content that could deceive the public.
The incident involves the intentional creation and dissemination of an AI-generated deepfake of a country's leader to influence public opinion and political processes surrounding a food safety scandal.
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 political campaigners intentionally using a deployed AI voice cloning tool to generate a deepfake of the president's voice for political campaigning.
Added categories peak at 2: Minor (Epistemic Harm, Direct)
A political campaign in Taiwan used AI voice cloning to impersonate President William Lai criticizing the government, sparking police investigations and political disputes. While presenting minor national security risks, the incident highlights the evolving threat of deepfakes to democratic norms, domestic political stability, and civil liberties.
The AI system was used as a tool to perform the specific function of voice generation based on direct inputs, without exhibiting autonomous planning or agency.
Risk Level 3: Limited Risk. The report describes the use of AI-generated deepfakes, which are subject to transparency and mandatory labeling obligations under the EU AI Act.
The AI system's primary function in this incident was to clone and generate the voice of President William Lai to make it sound like he was criticizing the government.
The reports provide consistent and detailed coverage of the incident from multiple news sources, outlining the creation of the video, the police response, the political reactions, and the specific legal complaints filed by both sides.
The incident could be interpreted primarily as a political free speech dispute rather than an AI safety failure, though the central role of deepfake technology makes it an AI safety issue.
Technology
Realistic AI Voice Cloning
AI cloned the president's voice, making it sound highly authentic.
High-Fidelity Deepfake Tools
Advanced AI tools allow easy generation of realistic fake audio.
Data Inputs
Use of Public Audio for Training
Public voice recordings were used without consent to train the AI model.
Human Factors
Intent to Mislead Viewers
Creators used AI to impersonate a leader to criticize the government.
Low Public AI Literacy
General public struggles to distinguish cloned voices from real statements.
Process and Methods
Inconspicuous AI Disclaimers
Disclaimers were hidden in the description box rather than on the video.
Inadequate Content Verification
No robust process existed to verify the authenticity of political videos.
Management
Commissioning of AI Content
Political managers commissioned AI voice clones without safety oversight.
Lack of Ethical AI Guidelines
Campaign teams lacked clear guidelines on the ethical use of AI.
Regulatory Environment
Inadequate AI Basic Act
Current framework lacks clear legal liabilities and penalty mechanisms.
Narrow Deepfake Laws
Deepfake regulations are mostly limited to election-related fraud.
No Mandatory Labeling Rules
No legal mandate exists to prominently label AI-generated political media.
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.
