This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
On May 22, 2023, a fake, AI-generated image of an explosion at the Pentagon circulated on Twitter, shared by several verified accounts including one impersonating Bloomberg News. The hoax caused public confusion and a brief, measurable decline in U.S. stock market indices before authorities debunked the report. The incident highlighted the risks of AI-generated misinformation and the vulnerabilities of social media verification systems.
A Twitter/X account allegedly impersonating Bloomberg reportedly posted an image falsely showing an explosion near the Pentagon. Analysts reportedly described the image as likely AI-generated. The post reportedly spread through major accounts before officials confirmed no incident occurred. Markets reportedly dipped during the short period when the hoax circulated.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Using AI systems to gain a personal advantage over others such as through cheating, fraud, scams, blackmail or targeted manipulation of beliefs or behavior. Examples include AI-facilitated plagiarism for research or education, impersonating a trusted or fake individual for illegitimate financial benefit, or creating humiliating or sexual imagery.
Also classified as
The AI-generated image directly spread false information about a terrorist-style attack, leading to widespread public confusion.
The incident involved the creation of a deceptive deepfake image of a Pentagon explosion to intentionally manipulate public belief and financial markets for potential personal or political advantage.
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 disseminating a fake AI image to deceive the public and manipulate markets.
Added categories peak at 4: Severe (Epistemic Harm, Indirect)
An AI-generated image of an explosion at the Pentagon was shared by verified accounts on Twitter, causing brief public panic and a temporary drop in major U.S. stock indices before being debunked. The incident highlights how synthetic media can be weaponized to manipulate financial markets and spoof national security institutions.
The AI system (an image generator like Midjourney or Stable Diffusion) operated as a simple tool, performing the specific function of generating an image based on direct user prompts without autonomous planning or context retention.
Limited Risk: The incident involves AI-generated content (deepfakes/synthetic images), which falls under Risk Level 3 due to transparency obligations requiring users to be informed when media is AI-generated.
The AI system was used to generate a synthetic image of an explosion, mimicking a real-world photograph to deceive viewers.
The reports provide consistent, detailed, and cross-verified accounts of the timeline, the specific platforms involved, the market reaction, and the official responses from the Pentagon and local emergency services.
None. The consensus across all reports is that this was a synthetic media hoax that successfully manipulated financial markets.
Technology
Generative AI Image Synthesis
Tools like Stable Diffusion and Midjourney allowed creation of realistic hoaxes.
Algorithmic Amplification
Platform algorithms naturally prioritize and reward sensational content.
Automated Verification
Twitter automated its verification, replacing manual identity checks.
Data Inputs
Lack of Source Verification
Accounts reshared the fake image without verifying its original source.
Missing Image Metadata
No digital signatures or metadata flagged the image as AI-generated.
Human Factors
Confirmation Bias
Users let down their guard and shared the image because it felt right.
User Impersonation
Bad actors paid for blue checks to impersonate trusted news sources.
Lack of Digital Literacy
Public and media struggled to distinguish paid checks from official ones.
Process and Methods
Gutted Content Moderation
Human moderation and fact-checking teams were laid off prior to the event.
Slow Institutional Debunking
Official verification channels moved slower than the viral spread.
Inadequate Media Vetting
News outlets and aggregators shared the image in a fast-paced environment.
Management
Prioritizing Revenue over Safety
Twitter prioritized subscription monetization over verification integrity.
Inadequate Risk Assessment
Platform leadership failed to foresee risks of pay-to-verify policies.
Regulatory Environment
Absence of AI Regulation
No comprehensive legal framework or licensing existed for advanced AI.
No Platform Safety Mandates
Lack of laws requiring platforms to filter synthetic misinformation.
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.
