This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
Following the fatal shooting of a nurse in Minneapolis, social media users circulated AI-altered images to distort the facts of the incident. One image, processed using Google's Gemini, was edited to falsely depict the victim pointing a gun at an agent and to remove a weapon from the agent's hand. These manipulations were used to support partisan narratives and smear the victim, complicating public understanding of the event.
Following the fatal shooting of Minneapolis ICU nurse Alex Pretti by U.S. Customs and Border Patrol agents, social media accounts reportedly circulated images purported to have been altered by AI, reportedly distorting evidence of the incident by portraying Pretti as threatening law enforcement and altering the presence of weapons. The images reportedly misidentified individuals and helped reinforce partisan narratives by purportedly obscuring verified video and eyewitness accounts.
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 Gemini AI tool generated false details, such as removing a gun and altering faces, when used to enhance the image, creating misleading content.
The incident involved the coordinated use of AI-altered images by high-profile social media influencers to conduct a disinformation campaign aimed at manipulating public opinion regarding a federal shooting.
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 was driven by human actors intentionally using deployed AI tools to alter images and spread disinformation to manipulate public opinion.
Added categories peak at 4: Severe (Epistemic Harm, Inferred)
Following a fatal shooting by federal agents in Minneapolis, domestic social media influencers used AI-altered images to spread disinformation and support partisan narratives. While the incident highlights the evolving threat of AI-enabled epistemic harm and public opinion manipulation, its direct national security impact remains minor and manageable under standard procedures.
The AI system, Google Gemini, was used as a tool to perform the specific, well-defined function of enhancing and sharpening an image based on direct user input.
Risk Level 3. Limited Risk: The report describes the use of AI-generated content and deepfakes (altered images) which pose a moderate risk and fall under transparency obligations.
The AI system was used to enhance, sharpen, and edit an existing image of the shooting scene, which falls under image retouching and upscaling.
The report clearly identifies the AI system (Google Gemini), its developer (Google), and the specific way it was used to alter an image of a high-profile event. The consequences of the alteration (removing a gun, changing faces) and the context of the disinformation campaign are explicitly detailed.
The image alterations could be viewed as simple user-driven photo editing rather than an AI safety failure, though the AI's generation of false details (removing the gun) represents a clear robustness/hallucination issue.
Technology
AI Image Hallucination
Gemini altered key details, removing a gun and changing faces in the image.
Generative Image Manipulation
AI tools allowed users to edit a phone to look like a gun pointed at an agent.
Data Inputs
Low Quality Source Footage
Shaky, low-resolution video inputs led to errors when AI attempted to sharpen.
Lack of Contextual Metadata
AI tools processed images without metadata to verify real-world authenticity.
Human Factors
Overreliance on AI Enhancements
Users trusted AI sharpening tools to reveal truth, despite inherent errors.
Confirmation Bias in Sharing
Influencers shared AI-altered images to support pre-existing narratives.
Process and Methods
Inadequate Verification Processes
Social media platforms lacked automated tools to flag AI-altered images.
Lack of Provenance Tracking
No standards existed to track if images had been modified by AI tools.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
