This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
Senator Ben Cardin was targeted in a sophisticated deepfake operation where a malign actor impersonated former Ukrainian Foreign Minister Dmytro Kuleba during a Zoom call. The impersonator used AI-generated voice and video to attempt to bait the Senator into commenting on sensitive political and foreign policy issues, including the use of long-range missiles in Russia. The Senator grew suspicious, ended the call, and reported the incident to authorities, who confirmed the impersonation. The event has prompted security warnings to other lawmakers regarding the increasing use of AI in social engineering and election interference.
Senator Ben Cardin was reportedly targeted with a purported deepfake on a Zoom call that appeared to imitate former Ukrainian Foreign Minister Dmytro Kuleba. The allegedly AI-generated video reproduced his likeness and voice but drew attention when the caller asked unusual questions.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Using AI systems to gain a personal advantage over others such as through cheating, fraud, scams, blackmail or targeted manipulation of beliefs or behavior. Examples include AI-facilitated plagiarism for research or education, impersonating a trusted or fake individual for illegitimate financial benefit, or creating humiliating or sexual imagery.
Also classified as
The operation was part of a broader, coordinated social engineering campaign by foreign actors aimed at influencing U.S. politics and elections.
The incident involved a highly targeted impersonation of a trusted foreign official using real-time deepfake technology to manipulate and bait a specific victim, Senator Ben Cardin.
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 a human actor who intentionally deployed deepfake technology to impersonate a foreign official and deceive a U.S. Senator.
Added categories peak at 3: Substantial (Epistemic Harm, Inferred)
Senator Ben Cardin was targeted in a sophisticated social engineering attempt using real-time deepfake technology to impersonate former Ukrainian Foreign Minister Dmytro Kuleba. The incident represents an evolved capability in information warfare and a direct threat to government functions, prompting security warnings across Congress.
The reports describe the deepfake technology as a tool used by a human 'malign actor' to alter their voice and appearance in real-time during a Zoom call. The AI system did not act autonomously or make independent decisions; it functioned purely to execute the specific task of mapping the actor's inputs to the likeness of Dmytro Kuleba.
Risk Level 3: Limited Risk. The incident involves the use of 'AI-generated content such as deepfakes' to impersonate a foreign official. Under the EU AI Act, deepfakes are classified as limited risk, requiring transparency obligations to inform users that the content is artificially generated or manipulated.
The AI system was used to generate a real-time deepfake video of Dmytro Kuleba's face, while also synthesizing or altering the actor's voice to match Kuleba's.
The reports are highly consistent, detailed, and corroborated by official Senate security notices, Senator Cardin's own statements, and multiple major news outlets. The facts of the Zoom call, the impersonation of Dmytro Kuleba, the specific questions asked, and the subsequent security warnings are clearly documented without conflicting accounts.
None. The event is clearly a targeted social engineering attempt leveraging real-time deepfake technology.
Technology
Live Deepfake Piping Tools
AI tools allowed seamless integration of live video and audio deepfakes.
Decreased Technical Barriers
Recent leaps in AI make sophisticated deepfakes easier to conduct.
High Technical Believability
The deepfake matched the target appearance and voice with high realism.
Data Inputs
Public Media of Target
Abundant media of Kuleba enabled highly accurate AI voice and face training.
Spoofed Email Request
A deceptive email initiated the contact, masquerading as a trusted source.
Relationship Intelligence
Attackers used known relationship details to make the contact look legitimate.
Human Factors
Initial Trust in Visuals
Staff initially trusted the call because the face and voice matched Kuleba.
Behavioral Suspicion
The senator grew suspicious due to out-of-character, aggressive questions.
Process and Methods
Lack of Real-Time Detection
There are no foolproof methods to detect deepfakes during a live call.
Weak Identity Verification
The office failed to verify the caller identity prior to starting the Zoom.
Active Social Engineering
Threat actors leverage existing relationships to bypass traditional defenses.
Management
Inadequate Verification Protocols
Senate offices lacked strict pre-verification steps for virtual meetings.
Need for Awareness Training
A lack of timely deepfake recognition training left staff vulnerable.
Regulatory Environment
Lack of AI Distribution Bans
The availability of deepfake creation tools remains largely unregulated.
Foreign State Interference
Malign foreign actors exploit AI to target democratic processes.
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.
