AI Multimodal Face Swapping and Voice Changing CEO Impersonation Transfer Scam: Extracted Voiceprints and Forged Videos Used in Emergency Directives to Defraud Corporate Funds

Who Gets Targeted

The primary victim demographics are corporate finance personnel, middle-to-senior managers, and sole proprietors with urgent capital allocation needs. Victims commonly exhibit a psychological tendency to obey authoritative superiors, easily falling into cognitive tunnels when facing high-pressure scenarios such as "urgent remittances" or "bidding deposits." Simultaneously lacking awareness of the breakthrough speed of AI deepfake technology, they tend to rely on facial and voice recognition as their sole identity verification methods, making their psychological vulnerabilities precisely exploitable.

骗局怎么运作

• Non-contact collection of voiceprint and facial materials: Scam syndicates make "silent phone calls" to induce victims to respond with "Hello, who is this?", capturing several seconds of their voiceprint features. Concurrently, they crawl facial images and public speech video clips of target executives from social media, corporate websites, or short-video platforms, using low-cost open-source tools for model training to build a basic forgery material library.
• Customized multimodal forgery generation: Utilizing AI face-swapping tools costing only 20 yuan and open-source voice cloning models, fraudsters produce real-time face-swapping videos that closely match the target leader's face, synthesize speech with specific accents and tones, and even simulate catchphrases and coughing sounds to bypass the victim's instinctive judgment of authenticity.
• Establishment of false communication channels and situational pressure: By sending SMS messages or emails with attached phishing links, fraudsters induce finance personnel to join hijacked office software or forged video conference rooms. During the conference, scammers use face-swapping synthetic videos to play the "superior," demanding that cameras be turned off for mutual communication under pretexts such as "commercial secrets" or "urgent bidding needs," thereby constructing an extremely urgent remittance scenario.
• Inducement to execute urgent, large-scale transfer directives: In the forged video or voice call, the "fake leader" uses an extremely stern tone to demand that finance personnel immediately transfer large sums of money to designated accounts, claiming it involves major M&A or bidding deposits that must be executed immediately, while threatening that delays will cause the company to miss golden opportunities or face accountability, thereby leveraging authority-obedience psychology to dismantle the victim's defenses.
• Rapid money laundering and cross-border fund transfer: Once funds are transferred into designated accounts, the scam syndicate immediately splits and transfers them overseas via aggregation platforms, cryptocurrency purchases, or underground banks, completing the laundering process within minutes. Simultaneously, all communication channels are severed. By the time victims verify the situation with real leaders through traditional channels, the golden window to stop payment has often passed, making fund recovery extremely difficult.

红旗信号（看到这些快跑）

• 🚩 Technical flaws generated by AI during calls or video conferences, such as video stuttering, blurred edges, and slight desynchronization between lip movements and sound.
• 🚩 The counterparty suddenly requests large fund transfers under various pretexts, while extremely emphasizing urgency and confidentiality and forbidding verification with anyone else.
• 🚩 The caller ID number or social media account differs from usual, yet the counterparty's voice and appearance are highly similar and extremely deceptive.
• 🚩 During a video conference, a leader suddenly requests a transfer to an unfamiliar account, often mentioning "bidding deposits" or new business not previously encountered.
• 🚩 Receiving a silent call or an unfamiliar incoming call with only background noise, followed shortly afterward by identity impersonation and inducement behaviors.

真实案例

• In 2024, a technology company branch head (Person A) engaged in a many-to-many video call with several "corporate executives" and "colleagues" during a so-called "headquarters video conference." Because the video feed and voices were indistinguishable from real people, Person A was instructed during the meeting to transfer funds to 5 designated local bank accounts, totaling HKD 200 million defrauded.
• In 2023, a female corporate finance worker in a certain region received a WeChat video call from someone claiming to be her "company boss," showing the boss's face and voice. Claiming an urgent need to pay project funds, the boss demanded an immediate transfer of RMB 4.3 million to an unfamiliar account. The woman completed the transfer within 10 minutes and only realized she had been scammed after verifying with the boss in person afterward.
• Public security authorities reported that in 2023, a corporate executive (Person B) received a help call from a "friend" whose voice was entirely identical to the real friend, requesting a loan of RMB 300,000 for temporary turnover. After transferring the funds, Person B contacted the friend to verify the situation and realized they had encountered an AI voice cloning scam, with cumulative losses reaching millions of yuan.

Official Stance

• In February 2024, the Cyber Security and Technology Crime Bureau of the Hong Kong Police Force issued a warning pointing out that fraud syndicates were utilizing AI deepfake technology to create many-to-many fake video conferences, reminding enterprises to strengthen verification mechanisms.
• In 2023, the National Anti-Fraud Center issued a prevention warning reminding the general public to be vigilant against new types of telecom fraud such as "AI face-swapping and voice-changing," emphasizing not to blindly trust portraits and voices in videos.
• In 2024, the Cyber Security Bureau of the Ministry of Public Security issued prevention notices pointing out that AI voiceprint and facial data are becoming key targets for black market theft, requiring enterprises to strengthen financial approval and internal information security.

How to Protect Yourself

• ✅ Establish internal multiple transfer approval mechanisms within enterprises, enforcing at least a dual-signature and phone verification process for any unconventional large-scale capital allocations, strictly prohibiting single-person direct transfers based solely on instant messaging directives.
• ✅ Request the counterparty to perform specific actions during video calls, such as waving, covering the face, or turning the head, to observe whether there are AI face-swapping flaws like facial distortion or video freezing, thereby increasing the difficulty of interactive confirmation.
• ✅ Protect personal biometric information by avoiding the casual publication of high-definition videos and audio on public platforms, maintaining vigilance against silent or abnormal unfamiliar calls, and preventing voiceprint features from leaking over the phone.
• ✅ Deploy watermarking and encryption technologies to protect corporate internal communication systems, and regularly conduct specialized anti-deepfake attack safety training and simulation drills for finance personnel and relevant executives.