This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
During the Iran war, a network of anonymous accounts on X utilized AI-generated videos and images to spread disinformation, specifically linking U.S. and Israeli military actions to Epstein-related conspiracy theories. This campaign, which included antisemitic content, aimed to erode public support for military operations and influence global opinion. Researchers identified the network as part of a pro-Iran propaganda effort, noting that the AI-generated content was used to lend false credibility to these narratives.
During the Iran war, anonymous pro-Iran accounts on X including HDX News and GPX News reportedly circulated a purportedly AI-generated Jeffrey Epstein-themed video while claiming it was real and that U.S.-Israeli strikes on Iran were launched to distract from Epstein-related material. The Washington Post reported the HDX News post received more than 6.8 million views and GPX News's version received more than 4.7 million before both accounts were suspended by X.
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 campaign generated and spread explicitly antisemitic and pro-Nazi content, exposing millions of social media users to hate speech.
The incident involves a coordinated, state-aligned network using AI-generated deepfakes to conduct a large-scale disinformation campaign aimed at manipulating public opinion and political processes.
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 operators intentionally deploying AI-generated deepfakes as part of a coordinated propaganda campaign.
Added categories peak at 4: Severe (Epistemic Harm, Direct)
A coordinated pro-Iran propaganda network utilized AI-generated deepfakes to conduct a high-impact information warfare campaign targeting the US and Israel during active military conflict. The campaign spread viral conspiracy theories and antisemitic content to erode public support for military actions, highlighting the evolving national security threat of AI-enabled disinformation.
The AI systems were used as simple generative tools by human operators to create fake videos and images. The reports do not describe any autonomous decision-making, planning, or multi-agent coordination by the AI itself.
Limited Risk: The incident involves 'AI-generated content such as deepfakes' used to spread disinformation. The report describes 'AI-generated fakes' and 'AI-generated videos' being spread on social media platforms without disclosure.
The AI was primarily used to generate highly realistic fake videos (deepfakes) of public figures to spread on social media platforms.
The reports provide detailed, credible accounts from disinformation researchers (ISD, ADL, Atlantic Council) regarding the coordinated campaign, specific view counts, and the nature of the AI-generated content. The role of AI as a tool for generating deepfakes is clear, and the societal impacts are well-documented.
The campaign could be viewed primarily as a geopolitical propaganda effort where AI was incidental, rather than an AI-specific safety failure.
Technology
AI Deepfake Generation Tools
Generative AI enabled creation of highly realistic fake videos and images.
Algorithmic Amplification
Recommendation algorithms prioritized engaging, controversial AI content.
Lack of Automated Detection
X lacked automated tools to instantly flag or block AI-generated fakes.
Data Inputs
Real Footage Mixed with AI
Combining real video with AI-generated scenes made disinformation look real.
Unlabeled AI Media
Lack of metadata or watermarks on AI content made tracking difficult.
Human Factors
User Confirmation Bias
Users shared AI fakes because the content aligned with existing beliefs.
Intentional Malicious Sharing
Influencers and state actors knowingly spread AI fakes to drive engagement.
Low Digital Literacy of Users
General public struggled to distinguish AI-generated deepfakes from reality.
Process and Methods
Slow Moderation Workflows
Platforms relied on delayed manual reviews and reports to remove AI fakes.
Inadequate Verification Process
Verification badge system allowed anonymous propaganda accounts to buy reach.
Delayed Policy Enforcement
New rules for AI-generated war videos were announced after viral spread.
Management
Monetization of Engagement
Platform incentivized sensational AI content via creator revenue sharing.
Insufficient Safety Investment
Platform prioritized rapid feature rollout over robust AI safety guardrails.
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.
