This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
A state-sponsored disinformation campaign in Venezuela utilized Synthesia's AI avatar technology to create fake news reports featuring fictitious anchors. These videos, which promoted government narratives regarding the economy and opposition, were widely distributed on social media and state-run television. The incident highlights the use of generative AI as a tool for sophisticated propaganda and the challenges of misinformation in environments with restricted access to independent news.
Video featuring fictitious news anchors was created using Synthesia to allegedly spread disinformation about Venezuela's economy on social media and Venezuelan state-run broadcast.
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 avatars were used to generate specific false narratives and misleading economic data regarding Venezuela's carnival tourism and baseball revenues.
The incident involves a coordinated, state-sponsored campaign using AI-generated avatars to spread political disinformation and manipulate public opinion at scale in Venezuela.
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 (the Venezuelan government and its supporters) intentionally using Synthesia's deployed AI platform to generate and spread political propaganda.
Added categories peak at 4: Severe (Epistemic Harm, Inferred)
The Venezuelan government utilized commercial generative AI technology to create synthetic news anchors that broadcast state-aligned propaganda. This incident represents a significant evolution in state-sponsored information warfare, leveraging deepfake technology to manipulate domestic and international public opinion in a highly restricted media environment.
The Synthesia AI system operated as a tool, performing the specific function of generating video avatars based on direct user-provided text scripts without autonomous planning or context retention.
Limited Risk: The report describes the creation of deepfakes and AI-generated video content, which under the EU AI Act are subject to specific transparency obligations to ensure users are informed that the content is synthetic.
The primary function of the Synthesia platform in this incident was to generate realistic deepfake video avatars that spoke user-provided scripts with synthesized voices.
The reports provide consistent, detailed accounts of the Venezuelan government's use of Synthesia's software to create fake news videos, including the specific names of the avatars, the platforms used for distribution, and the response from the AI developer.
None
Technology
Generative AI Deep Learning
Deep learning algorithms generated highly convincing talking human avatars.
Realistic Text to Speech
Synthetic accents and voices were perfectly synchronized with avatar mouth movements.
Lack of Built-in Watermarks
Initial software versions lacked digital watermarks to identify content as AI.
Data Inputs
User-Generated Scripts
Malicious actors inputted custom written scripts containing fake news narratives.
Real Actor Digital Twins
The catalog of avatars used images of real actors, raising perceived realism.
Human Factors
Target Audience Susceptibility
Censored populations lacked access to reliable news to verify the stories.
Malicious Intent of Clients
Chavismo actors systematically exploited the tool to produce propaganda.
Difficulty in Manual Detection
Average users could not easily spot subtle errors in the realistic avatars.
Process and Methods
Insufficient Screening Process
Internal screening failed to prevent bad actors from registering and using tools.
Small Content Moderation Team
The moderation team was too small to review the volume of videos generated.
Delayed Video Removal
Videos were only flagged after they had already gone viral on social media.
Management
Inadequate Risk Mitigation
Company did not anticipate political misuse of realistic news-anchor avatars.
Permissive Initial Policies
Restrictions on news-style content were only implemented after the incident.
Regulatory Environment
Widespread Media Censorship
State-controlled blocking of independent news created an information desert.
Lack of Deepfake Regulations
No legal frameworks existed to restrict synthetic state-sponsored propaganda.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
