This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
A deepfake video circulating on Instagram manipulated footage of a CBS News Texas anchor and fabricated an image of a stabbing suspect holding a knife. The incident occurred following a fatal stabbing at a Frisco, Texas, high school track meet, contributing to the spread of misinformation. The original post was removed, but the event highlights the ongoing challenges of AI-generated disinformation on social media platforms.
After a fatal stabbing at a Frisco, Texas, high school track meet, online accounts reportedly circulated a purported deepfake video on Instagram that allegedly both altered CBS anchor Doug Dunbar's speech and falsely depicted suspect Karmelo Anthony holding a knife. The manipulated content was reportedly reposted across multiple accounts, spreading misinformation about the case and intensifying inflammatory online discourse after the killing.
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 deepfake was used by race baiters to spread racially charged, toxic misinformation surrounding a violent crime.
The deepfake video impersonated a trusted news anchor and fabricated a suspect's likeness to manipulate viewer beliefs.
The incident involves online trolls and race baiters intentionally using AI-generated deepfakes to spread coordinated misinformation and manipulate public opinion following a violent crime.
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 deepfake videos were intentionally created and spread by human actors (trolls and race baiters) using post-deployment AI generation tools to manipulate public perception.
Added categories peak at 2: Minor (Epistemic Harm, Direct)
An AI-generated deepfake manipulated a CBS News Texas anchor and fabricated an image of a stabbing suspect to spread racially charged misinformation on Instagram. While the incident caused local concern and highlighted social media moderation challenges, its national security impact remains minor and localized, with no foreign state involvement or critical infrastructure threats.
The AI system functioned as a tool, performing the specific task of generating synthetic video and images from user inputs (such as a single still image) without exhibiting independent planning or autonomous decision-making.
Limited Risk: The report describes 'AI-generated deepfake videos' which fall under the category of AI-generated content requiring specific transparency obligations to ensure users are informed of their synthetic nature.
The report explicitly describes the creation of AI-generated deepfake videos, including a manipulated video of a news anchor and a fabricated image of a suspect holding a knife.
The report clearly details the nature of the deepfake, the specific individuals targeted (the anchor and the suspect), the platform used (Instagram), and the broader context of social media moderation policies. There is little ambiguity regarding the intent of the creators or the mechanism of the AI's involvement.
None. The incident is a straightforward case of malicious deepfake generation for disinformation.
While the specific AI model name is missing, this does not prevent reliable classification under the taxonomies.
Technology
Rapid Deepfake Tech Advancement
AI advances make high-quality deepfakes easy to create with one still image.
Convincing AI Video Generation
AI algorithms generate highly realistic and convincing manipulated footage.
Data Inputs
Small Dataset Requirements
AI models now require only a single still image to generate realistic deepfakes.
Human Factors
Malicious Intent of Online Trolls
Trolls and race baiters intentionally spread misinformation using deepfakes.
Process and Methods
Relying on User-Flagged Community Notes
Shifting moderation to users weakens proactive detection of deepfakes.
Removal of Professional Fact-Checkers
Ending active fact-checking processes allows faster spread of misinformation.
Management
Meta Policy Change on Fact-Checking
Management ended the use of professional fact-checkers on platforms.
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.
