Silent AI Call Scam: No Speaking After Answering, Using Two 'Hello's to Capture Your Voiceprint for Cloning Fraud
The main victim groups are the elderly, middle-aged and older people unfamiliar with smart devices, and hurried office workers who tend to speak first and ask questions on the phone. When these people receive calls from unknown numbers, they often unintentionally provide voiceprint samples through the polite response 'Hello? Who is this?' Their psychological weaknesses are: first, a natural trust in 'official numbers' and 'real human voices'; second, worry about account security or the safety of relatives and friends, making them prone to acting in haste without verifying. Scammers exploit this mental inertia, using silence to create suspense, then using realistic cloned voices to break through psychological defenses.
Key Fields
FIELD STAMPSWho Gets Targeted
The main victim groups are the elderly, middle-aged and older people unfamiliar with smart devices, and hurried office workers who tend to speak first and ask questions on the phone. When these people receive calls from unknown numbers, they often unintentionally provide voiceprint samples through the polite response 'Hello? Who is this?' Their psychological weaknesses are: first, a natural trust in 'official numbers' and 'real human voices'; second, worry about account security or the safety of relatives and friends, making them prone to acting in haste without verifying. Scammers exploit this mental inertia, using silence to create suspense, then using realistic cloned voices to break through psychological defenses.
骗局怎么运作
- Step 1: Mass outbound calling and silent probing: Fraud gangs rent AI outbound calling systems and automatically dial thousands of numbers every day. After the call is answered, the AI does not get straight to the point; instead, it stays silent for 1 to 3 seconds or plays a noise similar to a poor signal. At this point, many people who answer habitually say 'Hello? Is anyone there?' and this one sentence becomes the starting point for voiceprint collection.
- Step 2: Voiceprint collection and feature modeling: The backend AI records voice clips from the call in real time, extracts biometric features such as timbre, speaking rate, and waveform, and uses deep learning algorithms to build a digital voiceprint model of the speaker. To collect more samples, the AI repeatedly asks, 'The signal is bad, could you say that again?' guiding victims to repeat short phrases such as 'Hello,' 'Yes,' and 'I agree.' Modeling can be completed in just over ten seconds.
- Step 3: Marking vulnerable groups and collecting information: During the interaction, the AI also uses natural language processing to identify the victim's age, gender, tone tension, etc., and automatically marks them in the database as 'easily defrauded' or 'high value.' Some systems simultaneously extract key information such as the victim's region and commonly used forms of address, making it easier to customize scripts later. These data are packaged and sold to downstream fraud gangs, forming a black-market chain.
- Step 4: Voiceprint cloning and forged calls: After obtaining the voiceprint model, the fraud gang uses a speech synthesis engine to generate a voice almost identical to the victim's own. They then use the cloned voice to call the victim's relatives and friends, impersonating the victim and saying emergency lines such as 'I was in a car accident' or 'I'm being held.' Because the voice is highly similar, relatives and friends find it hard to detect anything abnormal and may not even call back to verify.
- Step 5: Combining scripts for precision fraud: Scammers also use previously collected social information to accurately call the victim's nickname or kinship terms during the call, further increasing credibility. For example, they tell a parent, 'Mom, my friend is hospitalized and urgently needs surgery fees,' and use reasons such as 'my phone got water damage' or 'the signal is bad at the hospital' to prevent the other party from calling back. Once the victim believes it, they immediately guide a transfer to a designated account.
- Step 6: Money laundering and disappearing traces: After succeeding, the fraud gang quickly splits the funds into multiple secondary accounts and then launders them through cryptocurrency or overseas accounts. At the same time, they shut down the outbound calling lines and delete cloud call recordings and voiceprint files, leaving no trace in the entire chain. By the time the victim notices and calls the police, the funds have often already flowed overseas, making recovery extremely difficult.
红旗信号(看到这些快跑)
- 🚩 An unknown caller stays silent for several seconds after the call is answered, or repeatedly says 'I can't hear you' or 'the signal is bad,' but delays stating their purpose.
- 🚩 The incoming number shows as 'Unknown,' 'Overseas,' or similar to an official number but slightly different, and the caller claims to be customer service, police/prosecutors/courts, or a relative or friend.
- 🚩 The other party asks you during the call to 'say your name again' or 'repeat this sentence,' clearly trying deliberately to get you to speak.
- 🚩 The call background sound has an obvious mechanical quality or delay, and the responses often do not answer the question, as if a recording is being played.
- 🚩 Shortly after you hang up, your relatives or friends receive a call from 'you' asking to borrow money or for help, and the caller cannot confirm via video call.
真实案例
- In April 2026, according to an investigative report by Xinhua Net, AI 'customer service' phone scams appeared in many places. Scammers impersonated e-commerce platform customer service, using the pretext of 'refund claims' or 'abnormal account' to induce victims to reveal verification codes or download screen-sharing software, with single cases losing tens of thousands of yuan. The report pointed out that many of these calls were automatically dialed by AI systems, and after the call was answered, they would ask victims to 'read out the verification code,' which was actually collecting voice samples. (Source: https://www.news.cn/legal/20260428/ccc23e82b9574c7ca1a641d0d10bbf26/c.html)
- In July 2026, Fuzhou Minhou police reported a case of AI voice cloning fraud. An elderly person received a call from 'grandson,' whose voice was almost identical to the real grandson, claiming to have been detained at a police station for fighting and needing bail money. The elderly person believed it and went to the bank to transfer money, but bank staff noticed something abnormal and called the police. Police investigation found that the voice had been cloned by scammers using AI technology, and the case was successfully intercepted. (Source: http://gs.news.cn/20260429/e7776bbda02844849d57d4930189e0bf/c.html)
- In July 2026, according to a CCTV News investigation, AI outbound calls have become rampant, with some companies making thousands of marketing calls a day, among which scam calls are mixed in. The investigation found that some AI outbound calling systems offer a 'de-labeling' service that can bypass mobile security software's harassment alerts. Once consumers answer and speak, their voices may be collected for subsequent precision fraud.
Official Stance
- On April 28, 2026, Xinhua Net published 'Xinhua Investigation | Beware of AI 'Customer Service' Phone Scams,' reminding the public to stay alert to AI voice calls and not to casually disclose personal information or verification codes.
- On July 1, 2026, the Minhou County Public Security Bureau issued a warning via Fuzhou News Network, exposing the modus operandi of using AI-cloned voices to impersonate relatives in fraud, and reminding people to always verify by video before transferring money.
- In 2026, the Ministry of Public Security reported that in the first half of the year, more than 170 cases of using AI tools to generate online rumors were investigated and handled, including multiple cases of fraud using AI voice synthesis, and stressed that they will be severely cracked down on in accordance with the law.
How to Protect Yourself
- ✅ Enable 'harassment blocking' and 'silence unknown numbers' on your phone, so unknown calls neither ring nor pop up, reducing the probability of answering from the source.
- ✅ For any call claiming to be customer service, police/prosecutors/courts, or a relative or friend, especially when a transfer is involved, hang up first, then proactively call back to verify using the official app or a saved real number.
- ✅ Agree with family members on a 'safe word' for emergencies; if you hear a family member say the wrong code on the phone, hang up immediately and call the police.
- ✅ Turn off microphone permissions for infrequently used apps on your phone, and regularly check under 'Settings - Privacy - Microphone' to prevent background recordings from being exploited by the black market.
- ✅ Download the National Anti-Fraud Center App and enable its incoming call warning feature. Also inform older family members about new scams such as 'silent calls' to help them raise their awareness of prevention.