This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
In mid-June 2025, an unknown actor conducted a social engineering campaign using AI-generated voice clones and synthetic text to impersonate U.S. Secretary of State Marco Rubio. The perpetrator contacted at least five high-level officials, including foreign ministers and a U.S. governor, via Signal and text messages, attempting to gain access to sensitive information or accounts. The State Department confirmed the incident and issued a warning to diplomatic staff, noting that while no direct cyber breach of department systems occurred, the campaign exploited trust and social engineering tactics.
Sometime in mid-June 2025, an unknown actor reportedly used AI voice-cloning and text generation to impersonate U.S. Secretary of State Marco Rubio. The impostor allegedly contacted at least five high-level officials, including three foreign ministers, a U.S. governor, and a member of Congress, via Signal and text messages. The actions reportedly aimed to manipulate targets into sharing sensitive information or account access.
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 campaign represents a sophisticated espionage or influence operation targeting diplomatic relationships and state-level communication channels.
The attacker used AI voice cloning to impersonate Secretary of State Marco Rubio to target specific high-level officials in order to fraudulently gain access to their accounts or sensitive information.
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 initiated by an unknown human actor who intentionally used deployed AI voice-cloning tools to impersonate a government official for espionage or information access.
Added categories peak at 2: Minor (Epistemic Harm, Direct)
An unknown actor utilized advanced AI voice cloning and text replication to impersonate U.S. Secretary of State Marco Rubio, targeting high-level U.S. and foreign officials in an espionage attempt. While no systems were breached and no targets were deceived, the incident demonstrates an evolved threat to diplomatic communications and sovereignty, requiring coordinated security responses.
The AI software functioned as a tool, performing the specific task of voice synthesis and text style replication based on direct human inputs (audio samples and text scripts) without exhibiting independent planning or autonomous decision-making.
Limited Risk: The incident involves AI-generated content (deepfakes), which falls under Risk Level 3 due to transparency obligations to ensure users are aware they are interacting with synthetic media.
The AI system was primarily used to clone and generate synthetic audio mimicking Marco Rubio's voice, as well as replicating his writing style in text messages.
The reports provide consistent, detailed accounts of the impersonation campaign targeting Marco Rubio, citing official State Department cables and FBI warnings. While the identity of the attacker remains unknown, the technical mechanism (AI voice cloning and text synthesis) and the scope of the targets are clearly documented across multiple reputable sources.
The incident could be viewed purely as a traditional social engineering/phishing campaign, but the central role of synthetic voice cloning makes it a distinct AI safety incident.
Technology
Accessible Voice Cloning Tools
Advanced AI platforms require only seconds of audio to clone voices.
Lack of Real-time AI Detection
Traditional security tools cannot flag synthetic voice and text in real-time.
Realistic Writing Mimicry
Generative AI chatbots produce scripts mimicking specific writing styles.
Data Inputs
Abundant Public Audio Data
Hours of public speeches allow easy synthesis of the target's voice.
Leaked Private Contact Lists
Compromised phone numbers enabled precise targeting of high-level officials.
Spoofed State Domain Names
Impostor used spoofed domains and branding to increase credibility.
Human Factors
Exploitation of Human Trust
Attackers preyed on cognitive biases and urgency in high-pressure scenarios.
Carelessness About Data Security
Officials used insecure channels like Signal for official business.
Overreliance on Vocal Familiarity
Recipients trust familiar voices, bypassing standard verification.
Process and Methods
Inadequate Verification Protocols
Lack of structured processes to verify the source of voice and text messages.
Insecure Messaging App Usage
Using Signal without organizational markers or security oversight.
Flawed Authentication Methods
Traditional passwords and codes fail when users are tricked by cloned voices.
Management
Inadequate Security Oversight
Failure to enforce secure communication rules and monitor tool usage.
Broad IT System Access Grants
Sweeping access granted to external entities created latent vulnerabilities.
Hollowing Out Defensive Programs
Budget cuts projected to reduce workforce focused on deepfake threats.
Regulatory Environment
Gaps in Platform Moderation
AI services lack robust moderation to prevent unauthorized voice cloning.
Weak Legal and Policy Frameworks
Insufficient regulations and consequences for spreading AI disinformation.
Lack of Content Provenance Rules
No standards exist to verify the authenticity of digital media files.
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.
