AI Outbound Phone Scams: Deepfake Voiceprint Extortion of Victims

Who Gets Targeted

Victims are predominantly seniors over 60, alongside individual users such as small-to-medium enterprise owners and e-commerce sellers. Geographically, cases have spread from Gansu to multiple provinces including Jiangsu and Shandong. Believing they genuinely hold memberships, auto-renewal subscriptions, or pending refunds, victims readily trust AI-synthesized voices that are indistinguishable from official customer service. Prompted by anxieties manufactured through scripts like 'auto-deduction of 800 yuan per month' or 'account anomalies,' they easily fall into the trap of downloading unfamiliar apps, sharing their screens, or disclosing verification codes. Once a phone is remotely controlled, funds are quickly transferred and difficult to recover. Individual losses range from tens of thousands to hundreds of thousands of yuan. In Jinchang, Gansu alone, over 20 victims have been affected with total losses exceeding 3 million yuan (according to the Jinchang Municipal Public Security Bureau Jinchuan Branch briefing, reported by Xinhua News Agency on 2026-04-28).

骗局怎么运作

• Collect target voice samples: Fraud rings acquire recordings of victims or public speakers via social media or open channels, accumulating audio materials containing speech habits and details to prepare for subsequent model training.
• Train voiceprint cloning models: Using open-source or commercial deep learning frameworks, the collected audio is labeled and trained to produce synthesis models capable of reproducing the target voice with high precision.
• Generate customized scripts: Write dialogue scripts matching common scam scenarios (such as bank account anomalies, technical support, prize notifications) and have the voiceprint model generate real-time conversation content, ensuring high consistency between voice and text.
• Automatic outbound calling and number spoofing: Using VoIP platforms or bulk call services, batch-dial target numbers, employ disguise methods to conceal true carrier identifiers, and push calls featuring the same verified AI voice to form 'carpet-style marketing.'
• Social engineering payment and inducement: Rapidly manufacture urgent scenarios over the phone, demanding victims provide verification codes, bank card passwords, or transfer money to 'secure accounts,' leveraging the persuasiveness of familiar voices and official backgrounds to drastically increase transfer success rates.

红旗信号（看到这些快跑）

• 🚩 Phone numbers change frequently, originate from unknown sources, or appear as 'anonymous' or hidden numbers.
• 🚩 Voices are extremely similar to acquaintances or official personnel, exhibiting high-fidelity forgery traces.
• 🚩 Callers claim to be from banks, public security, government agencies, or official customer service, speaking in a formal and rigorous tone.
• 🚩 Urging immediate fund transfers or disclosure of personal verification codes, emphasizing that 'safety requires immediate action.'
• 🚩 Call duration is abnormally long, prompting victims to repeatedly verify information.

真实案例

• On April 28, 2026, Xinhua News Agency reported that the Jinchuan Branch of the Jinchang Municipal Public Security Bureau in Gansu Province cracked a telecommunications fraud case utilizing AI voice synthesis technology to impersonate official customer service, targeting senior citizens. After nearly half a month of reconnaissance and stakeouts, the task force arrested 8 suspects, verifying over 10 cases of the same type. Individual losses ranged from tens of thousands to hundreds of thousands of yuan, with total losses exceeding 3 million yuan. (Source: [https://www.news.cn/legal/20260428/ccc23e82b9574c7ca1a641d0d10bbf26/c.html](https://www.news.cn/legal/20260428/ccc23e82b9574c7ca1a641d0d10bbf26/c.html))
• On December 16, 2025, People's Daily reported that the People's Court of Huangshigang District, Huangshi City, Hubei Province, concluded a trial on an AI voice-imitation fraud case. The prime suspect, in collusion with overseas syndicates, used AI voice mimicry to impersonate a senior's grandson, defrauding three elderly individuals out of a total of 60,000 yuan under the pretext that the child had injured someone at school and required compensation. The defendant was sentenced to two years and one month in prison for the crime of fraud, along with a fine of 15,000 yuan; the judgment has taken effect.
• On December 17, 2025, Peninsula Metropolis Daily reported that an 80-plus-year-old resident in Beijing received a family scam call impersonating their son, losing 100,000 yuan in cash under the pretext of a car accident requiring urgent medical compensation. Beijing police arrested 8 suspects. The same report disclosed that Hubei Huangshi police consolidated and investigated 3 AI voice-imitation grandson fraud cases where the cash withdrawal person was the same individual, and the 60,000 yuan in illicit gains has been fully refunded to the three deceived seniors.

Official Stance

• The Financial Consumer Protection Bureau of the National Financial Regulatory Administration issued the 'Risk Warning on Preventing New Types of Telecommunications Network Fraud' on July 25, 2024, alerting the public to remain vigilant against 'AI face-swapping and voice-mimicry' scams (collecting pronunciations and synthesizing false audio to impersonate others, build trust, and induce transfers). (Source: [https://www.gov.cn/lianbo/bumen/202407/content_6964352.htm](https://www.gov.cn/lianbo/bumen/202407/content_6964352.htm))

How to Protect Yourself

• ✅ Confirm caller identity: After hanging up, directly call official customer service lines or visit bank counters to verify; never act directly on prompts given over the phone.
• ✅ Configure AI protection plugins: Install trusted anti-fraud software on mobile phones or computers, enabling 'AI outbound call identification' or 'call warning' features to timely recognize suspicious calls.
• ✅ Enhance information security awareness: Regularly update account passwords, enable two-factor authentication, and avoid casually leaving verification codes, passwords, and personal information in call logs.
• ✅ Report and retain evidence: Upon discovering suspected AI outbound call fraud, report it promptly via 12377 or official apps while retaining call recordings, communication logs, and transfer receipts to facilitate tracking and fund recovery.