A fabricated video spread online showing US Secretary of State Marco Rubio threatening to convince Elon Musk to cut off Ukraine's Starlink access, but the audio was AI-generated and the threat was never made in the actual CNN interview.
A video circulated on social media platforms including X (formerly Twitter) purporting to show US Secretary of State Marco Rubio in a CNN interview with Kaitlan Collins threatening to meet with Elon Musk to convince him to cut off Ukraine's access to Starlink internet service. The video began with authentic footage of Rubio responding to a question about what he wanted Ukrainian President Zelensky to apologize for, but then cut to footage of a Trump-Zelensky meeting while fabricated audio made it sound like Rubio was threatening Starlink access. The video spread across multiple platforms and languages including French and Polish. CNN confirmed the video was completely manipulated and did not air on their network. The State Department's official transcript of the February 28, 2025 interview shows Rubio made no mention of Starlink. Media forensics expert Siwei Lyu from the University at Buffalo analyzed the audio and determined it was very likely AI-generated, noting the clean sound without background noise or echo typical of AI-generated audio. Ukraine's Center for Countering Disinformation also confirmed the video was fake. The fabricated content emerged during a tense period when the US had suspended military aid and intelligence sharing with Ukraine following a contentious White House meeting.
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.
AI system
Due to a decision or action made by an AI system
Intentional
Due to an expected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed