Reported AI-Generated Video Call Impersonation of Cryptocurrency Analyst Leads to Alleged Malware Installation and Account Theft

Cryptocurrency analyst Mai Fujimoto reported losing access to her X, Telegram, and MetaMask accounts through a video call with a purported deepfake impersonating a trusted contact. According to Fujimoto, the attacker, who appeared on Zoom as her acquaintance, instructed her to install software under the guise of resolving an audio issue. Fujimoto believes this enabled malware installation. In response, Binance founder Changpeng Zhao warned such attacks render video call verification unreliable.

Risk classification

• Primary risk domain: 4 Malicious actors
• Primary risk subdomain: 4.3 Fraud, scams, and targeted manipulation

The incident involved hackers using AI deepfake technology to impersonate a trusted contact in a targeted video call to deceive the victim into installing malware, leading to account compromises.

Additional risk subdomains

• 5.1 Overreliance and unsafe use: The victim relied on the visual representation of her acquaintance during the video call despite persistent audio issues, leading to unsafe software installation.

Causal factors

• Entity: Human
• Intent: Intentional
• Timing: Post-deployment

The incident was caused by human hackers who intentionally deployed and used AI deepfake technology to execute a social engineering attack.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Risk Level 3: Limited Risk. The report describes the use of AI-generated content, specifically deepfakes, which falls under the category of systems posing moderate risk with transparency obligations.

AI system and alleged parties

• AI system: None named
• AI purpose: Deepfake Video Generation; Voice Generation
• Behaviour type: Tool
• Alleged developer: Unknown voice cloning technology developers, Unknown deepfake technology developers
• Alleged deployer: Unknown scammers, Unknown scammer impersonating acquaintance of Mai Fujimoto
• Alleged harmed parties: Unnamed acquaintance of Mai Fujimoto, Mai Fujimoto, General public

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Negligible. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Negligible
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Negligible, indirect Negligible
• Differential treatment: direct Negligible, indirect Negligible
• Civil rights: direct Negligible, indirect Negligible
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Negligible, indirect Minor
• Psychological: direct Negligible, indirect Negligible
• Epistemic: direct Negligible, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Privacy

Reported: The report explicitly describes privacy violations through the compromise of personal accounts.

Indirectly caused: The attack led to the compromise and unauthorized access of the victim's X, Telegram, and MetaMask accounts, violating her digital privacy.

Inferred additional harm: The hackers likely accessed private messages, personal data, and financial keys within the compromised accounts.

People affected

• Occurrences reported: 1
• People reportedly harmed: 1
• People reportedly exposed: 1

Potential causes

Management

• Weak Account Security Management: Initial compromise of the friend's account enabled the social engineering.

Technology

• Advanced AI Deepfake Synthesis: Real-time video synthesis allowed highly convincing impersonation of a friend.
• Malicious Link Execution: Installation of malware via a link shared during the video call.

Data Inputs

• Compromised Friend Account Data: Attackers used a compromised Telegram account to initiate trusted contact.
• Acquaintance Likeness Data: Attacker used the victim's friend's biometric and visual data for deepfake.

Human Factors

• Trust in Compromised Contact: Victim accepted invitation in good faith due to existing relationship.
• Cognitive Bias and Overtrust: Victim rationalized audio issues and followed malicious instructions.

Process and Methods

• Inadequate Video Verification: Lack of out-of-band verification to confirm the identity of the caller.
• Weak Software Install Protocols: Victim downloaded and executed untrusted software during a live call.

Regulatory Environment

• Lack of Deepfake Regulations: No strict legal frameworks preventing unauthorized biometric likeness synthesis.

Information quality

• Classification confidence: High
• Reason for confidence: The report clearly describes the social engineering attack, the use of deepfake technology during a Zoom call, and the resulting compromise of the victim's accounts. There is little ambiguity about the sequence of events or the role of the AI tool.
• Ambiguities identified: The specific deepfake software used and the exact financial value of assets lost from the MetaMask wallet are not specified.

A private cryptocurrency analyst fell victim to a social engineering attack utilizing real-time deepfake video technology during a Zoom call, leading to compromised personal accounts. While the incident is a localized cybercrime with minor national security impact, it highlights the rising threat of real-time synthetic media in bypassing traditional video verification methods.

• Overall national security impact: Minor
• Scope: Unknown
• Primary target: No clear primary
• Alleged perpetrator: Unknown

Threat characteristics

• Imminence: Long-term. The specific incident is resolved, but the underlying capability of real-time deepfake impersonation represents an ongoing strategic concern.
• Autonomy: Human-controlled. The AI deepfake technology was directly controlled and deployed by human hackers to facilitate a manual social engineering attack.
• Novelty: Evolved capability. Represents a significant advancement of existing deepfake capabilities, transitioning from static synthetic media to real-time interactive video impersonation.

Impact by dimension

• Physical security: Negligible. No physical systems, kinetic attacks, or critical infrastructure were affected by this individual social engineering incident.
• Information security: Minor. Demonstrates an evolved capability of real-time deepfake impersonation for credential theft, though currently limited to a single private individual.
• Sovereignty: Negligible. The attack targeted a private cryptocurrency analyst, with no impact on state sovereignty, electoral systems, or government operations.
• Economic security: Negligible. The compromise of a single analyst's cryptocurrency wallet and social media accounts does not threaten national financial systems, strategic technologies, or critical supply chains.
• Societal stability: Negligible. The incident was a localized cybercrime targeting one individual, presenting no threat to societal stability, civil liberties, or population-scale safety.