This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
The White House posted an image on X depicting a protester, Nekima [name withheld], appearing to cry during her arrest, which was identified as likely AI-manipulated. Analysis by The New York Times using Resemble.AI confirmed signs of manipulation, and the paper demonstrated that similar alterations could be produced using Gemini and Grok. The incident raises concerns regarding the potential impact on the legal prosecution of the protester.
The White House reportedly posted a purportedly AI-altered image on X showing Minnesota protester and attorney Nekima Levy Armstrong appearing to cry during her arrest. An earlier image reportedly shared by Homeland Security Secretary Kristi Noem showed her calm. Reported analysis by third parties using AI detection tools found signs of purported facial manipulation, which could be replicated using generative AI systems.
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 protester's skin was darkened and her facial expression altered to depict her crying, representing her unfairly and potentially prejudicing her legal case.
The White House used AI-manipulated media on official social media channels to conduct targeted political influence and shape public opinion regarding a high-profile arrest.
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 actors in the White House intentionally using generative AI tools post-deployment to alter and distribute a misleading image for political messaging.
Added categories peak at 2: Minor (Epistemic Harm, Direct)
The White House's distribution of an AI-manipulated image of a protester represents a minor national security concern. While it does not threaten physical infrastructure, it undermines information integrity from official channels and introduces legal complications that could hinder a federal prosecution, highlighting the risks of state-level adoption of synthetic media for political messaging.
The report describes generative AI systems being used as tools to perform specific image manipulation functions on direct input. The Times was able to replicate the image by asking Gemini and Grok to alter the original image.
Limited Risk: The incident involves 'AI-generated content such as deepfakes' which falls under Limited Risk due to transparency obligations to notify users when AI-generated content is used.
The AI systems (Gemini, Grok) were used to generate a deepfake image by altering a real photograph of an arrest.
The report clearly details the incident, the specific AI tools used to detect and replicate the manipulation, and the political and legal implications. There is little ambiguity regarding the facts of the image alteration.
None
Technology
Generative AI Image Modification
Generative AI tools enabled realistic alteration of the original arrest photo.
AI Face and Skin Manipulation
AI algorithms modified facial expressions and darkened skin tones easily.
Data Inputs
Real Arrest Photo as Input
Original authentic photo was fed into AI tools to generate the fake version.
Human Factors
Intentional Distribution of Memes
Officials chose to distribute AI-altered images for political communication.
Enthusiastic Sharing of AI Content
Key figures actively promoted synthetic and manipulated media.
Process and Methods
Lack of Media Verification
No formal process was used to verify the authenticity of official posts.
Management
Condoning Synthetic Media Use
Leadership defended the post, stating 'the memes will continue'.
Lack of AI Communication Policy
No management oversight to restrict the use of AI-manipulated images.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
