This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
The Russian disinformation campaign Storm-1516 used generative AI to create deepfake videos falsely accusing Texas Democratic U.S. Senate candidate James Talarico of sexual assault, impersonating a real former student to influence the election.
Two purported deepfake videos circulated on X using former student Zyon Lewis's identity to allegedly falsely accuse Texas U.S. Senate candidate James Talarico of sexual assault. Lewis said the person depicted was not him and denied the allegation. Researchers linked the videos to the Russia-aligned Storm-1516 influence operation. The original account was suspended, but copies continued to circulate.
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 system was used to generate highly damaging, completely fabricated claims of sexual assault to mislead voters.
The deepfake videos impersonated a specific, real individual (a former student) without his consent to damage the candidate's reputation.
The incident involves a coordinated, state-aligned foreign influence operation using AI-generated deepfakes to target a political candidate and manipulate public opinion during an election.
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 driven by human actors (the Storm-1516 group) who intentionally used deployed AI deepfake tools to create and spread political disinformation.
Added categories peak at 2: Minor (Epistemic Harm, Direct)
The incident involves a coordinated Russian state-aligned information operation targeting a U.S. Senate candidate with AI-generated deepfakes to influence an election. While the AI generation was low quality and quickly debunked, it represents a substantial threat to information security and national sovereignty by attempting to manipulate democratic electoral processes.
The report describes the AI system as being used to 'generate' or 'create' deepfake videos. It operated as a tool for the Russian propaganda group Storm-1516 to produce synthetic media, with no indication of autonomous decision-making or agentic behavior.
Limited Risk: The AI system was used to generate deepfakes (AI-generated content). According to the EU AI Act, providers of AI systems that generate or manipulate image, audio or video content (deepfakes) must disclose that the content has been artificially generated or manipulated.
The AI system was used to generate a synthetic video of a man speaking, which exhibited unnatural facial movements and unsynchronized audio, characteristic of deepfake video and voice generation.
The reports from Graphika and PolitiFact provide clear, consistent, and detailed evidence of a coordinated foreign influence operation using AI-generated deepfakes to target a political candidate. The roles of the actors, the platforms used, and the fact-checking process are well-documented, leaving little ambiguity about the nature of the incident.
None
While the specific AI model is not named, this does not prevent high-confidence classification of the incident's domain, causes, and harms.
Technology
Accessible Deepfake Tools
AI tools enabled synthesis of realistic video and voice to mimic a real person.
Flawed AI Generation Artifacts
Unnatural eye movements and stagnant expressions revealed the video was fake.
Weak Automated Detection
Platforms lacked automated tools to block deepfakes during upload.
Data Inputs
Harvested Personal Data
Public Facebook data of a real student was used to craft the fake persona.
Missing Provenance Metadata
Lack of cryptographic watermarks allowed anonymous AI video distribution.
Human Factors
Partisan Confirmation Bias
Conservative accounts reshared the video without verifying its authenticity.
Malicious Foreign Actors
Russian group Storm-1516 intentionally fabricated the sexual assault claim.
Impersonation Vulnerability
A real individual's identity was stolen to make the deepfake credible.
Process and Methods
Coordinated Amplification
Networks of bots and paid influencers quickly spread the deepfake video.
Reactive Platform Moderation
X suspended the account only after the video had already gone viral.
Weak Identity Verification
An account registered in Romania easily impersonated a Texas resident.
Management
Engagement-Driven Algorithms
Platform management prioritized viral reach over information integrity.
Poor Election Risk Planning
Failure to allocate sufficient resources to counter foreign AI campaigns.
Regulatory Environment
Lack of AI Election Laws
No federal regulations prohibiting AI-generated political disinformation.
Weak Platform Liability
Social media platforms face few legal penalties for hosting deepfakes.
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.
