This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
During Hurricane Helene in September 2024, AI-generated images depicting scenes of devastation, such as a child trapped in floodwaters, were widely circulated on social media. These images were used to spread misinformation, fuel partisan narratives, and potentially facilitate scams. The proliferation of this synthetic media caused 'disaster fatigue,' eroded public trust in legitimate news sources, and complicated emergency response efforts by creating confusion among the public and first responders.
During Hurricane Helene (September 24-29, 2024), purportedly AI-generated images circulated on social media, reportedly misleading the public and hindering disaster response efforts. Reportedly fake images including animals stranded on rooftops and political figures in floodwaters added confusion and disrupted emergency management. The spread of these images allegedly exacerbated existing challenges, including power outages and communication failures, all of which led to complications providing accurate information to those in need.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Highly personalized AI-generated misinformation creating “filter bubbles” where individuals only see what matches their existing beliefs, undermining shared reality, weakening social cohesion and political processes.
Also classified as
The synthetic images were actively used by political influencers and campaigns as propaganda to shape public opinion and polarize narratives regarding the government's disaster response.
The widespread flooding of social media with AI-generated disaster images led to 'verification fatigue' and public apathy, severely degrading the shared information ecosystem during a real-world crisis.
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 risk and subsequent harms were caused by human actors intentionally generating and sharing deceptive AI images to spread propaganda, gain social media engagement, or facilitate scams.
Added categories peak at 4: Severe (Epistemic Harm, Inferred)
The incident highlights the national security risks of generative AI being used to pollute the information ecosystem during a domestic natural disaster. While not a kinetic threat, the widespread dissemination of realistic deepfakes eroded public trust in government agencies, forced federal emergency responders to divert resources, and demonstrated an evolved capability for exploiting crises to drive societal polarization.
The AI systems were used as tools to generate static synthetic images based on direct user prompts, without demonstrating independent planning, context retention, or autonomous decision-making.
Risk Level 3: Limited Risk. The reports describe the generation and dissemination of 'deepfake' images and AI-generated content, which are subject to transparency obligations under the EU AI Act to ensure users are aware they are interacting with synthetic media.
The primary function of the AI systems involved was to generate highly realistic but fake images of disaster scenes (deepfakes) to be shared on social media platforms.
The reports provide consistent, detailed, and corroborating accounts of the specific viral AI-generated images, their visual flaws, the public figures who shared them, and the resulting societal and psychological impacts. There is high consensus across the sources regarding the nature of the incident.
None. The consensus across all reports is that the images are AI-generated fakes that caused public confusion.
Technology
Advanced Generative AI
Cheap and easy tools create highly realistic, deceptive synthetic images.
Algorithmic Amplification
Social media algorithms prioritize highly emotional and engaging fake content.
Physical Anomalies in AI Output
Generators produce subtle errors like extra fingers or mismatched colors.
Data Inputs
Lack of Image Provenance
Uploaded images lack verifiable metadata to trace their synthetic origin.
Exploitative Prompting
Users feed emotionally charged prompts to generate shocking disaster scenes.
Human Factors
Verification Fatigue
The mental effort to verify images leads users to disengage and stop caring.
Emotional Reactivity
Arresting visual images evoke strong emotions before cognitive processing.
Lack of Digital Literacy
Many users cannot identify subtle physical inconsistencies in synthetic media.
Process and Methods
Weakened Platform Moderation
Changes in verification and moderation allow unverified deepfakes to spread.
Inadequate Verification Speed
Manual fact-checking is too slow to counter the viral spread of AI images.
Erosion of Local News
Shrinking local news creates an information vacuum filled by fake AI content.
Management
Engagement Monetization Models
Platforms pay content farms based on engagement, incentivizing fake images.
Prioritizing Engagement Over Truth
Platform management altered algorithms to favor viral reach over accuracy.
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.
