This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
An AI-generated image circulated on social media falsely claiming to show hundreds of Tanzanians queuing for only five job openings at the Tanzania Revenue Authority. The image, which contained visual artifacts typical of AI generation, was used to misrepresent the actual job interview process and exacerbate public anxiety regarding youth unemployment. Detection tools from Sight Engine and Hive Moderation confirmed a high probability that the image was AI-generated.
An image circulated on Facebook purporting to show hundreds of Tanzanians lining up for just five jobs at the Tanzania Revenue Authority (TRA). The photo, reportedly AI-generated, gained traction shortly after TRA invited over 100,000 applicants for interviews on March 29–30. Though TRA had announced more than 1,500 openings, the image misrepresented the situation, stoking public anxiety during a period of high unemployment.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI systems that inadvertently generate or spread incorrect or deceptive information, which can lead to inaccurate beliefs in users and undermine their autonomy. Humans that make decisions based on false beliefs can experience physical, emotional or material harms
The incident involves the creation and viral dissemination of a single AI-generated image that spread false information regarding the number of job openings at a government agency.
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 actor intentionally generating and sharing a deceptive AI-generated image to spread misinformation, occurring after the deployment of the generative AI tool.
Added categories peak at 2: Minor (Epistemic Harm, Direct)
An AI-generated image was used to spread misinformation regarding recruitment at the Tanzania Revenue Authority, aiming to exploit public anxiety over youth unemployment. The incident represents a minor national security threat, primarily impacting information security and social stability at a localized level, with no disruption to critical infrastructure, government functions, or state sovereignty.
The generative AI system acted as a tool, performing the specific function of generating an image based on direct inputs without independent planning or autonomy.
Limited Risk: The incident involves AI-generated content (a deepfake image) which, under the EU AI Act, poses a moderate risk and requires transparency obligations to inform users that the content is AI-generated.
The AI system was used to generate a synthetic image depicting a crowd of people, which was then used as a deepfake to support a false narrative.
The report clearly details the nature of the misinformation, the specific false claims, the actual facts from the Tanzania Revenue Authority, and the technical verification of the image being AI-generated using reliable detection tools.
None
Technology
Generative AI Capabilities
AI tools enabled the creation of realistic but deceptive crowd images.
AI Generation Artifacts
The model produced physical distortions like missing ears and blurred faces.
Data Inputs
Missing Provenance Metadata
The image lacked embedded metadata or watermarks identifying it as AI.
Human Factors
User Confirmation Bias
Users shared the fake image because Tanzania's job crisis is a real issue.
Lack of Source Verification
Social media users reposted the image without checking its original source.
Process and Methods
Passive Platform Moderation
Social media platforms relied on reactive reporting instead of active filters.
Delayed Fact-Checking
The image was analyzed and flagged only after spreading widely online.
Regulatory Environment
Absence of AI Labeling Laws
No legal mandates exist requiring platforms to tag synthetic media.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
