Seeing a face on video or hearing a voice on the phone does not count as verification; if a transfer is involved, hang up first and call back using an old number saved in your contacts.

No cost involved. It only requires making one extra call and waiting a few extra minutes. In general, avoid posting videos of your face or long voice recordings on public accounts. The hard part is that scammers often pressure you to act quickly — you must resist the urge to transfer money right away.

The Cybersecurity Bureau of the Ministry of Public Security has outlined three key measures. First is motion verification: ask the other party to perform a smooth head-turn motion, or to completely cover their face with a palm and then remove it quickly. The original text notes that current AI technology still struggles to accurately simulate such complex facial occlusions and lighting changes, often resulting in choppy, blurry, or distorted images. Second is listening for audio cues: AI-generated voices sometimes lack natural breathing pauses or emotional variation, and may sound slightly mechanical or contain unusual background noise. Third and most important is cross-channel verification. As stated in the source material, “after hanging up, immediately call a number the person regularly uses and you recognize — do not call back the suspicious number — or ask a mutual acquaintance for confirmation.” The rule of thumb is: “Don’t trust video or voice alone; verify through actions and multiple channels.” Legally, using face-swapping or voice-cloning technology for fraud is strictly prohibited. Platforms must clearly label any service offering face generation, replacement, voice synthesis, or voice imitation; no organization or individual may remove, alter, or hide such labels. However, since scammers rarely follow these rules, the absence of a label cannot be taken as proof that the caller is genuine. The Criminal Investigation Bureau reported one case where a 14-member gang produced over 50,000 synthetic face videos and collected more than 500,000 pieces of personal information; police also seized 60 cloud servers nationwide. This labeling requirement took effect on September 1, 2025.

公安部网安局 (2025). 眼见≠为实，警惕 AI 新型诈骗！http://gat.fujian.gov.cn/ztzl/fjjffpzxrx/spjq/202509/t20250922_7013008.htm（福建省公安厅转载）；国家互联网信息办公室、工业和信息化部、公安部 (2022). 互联网信息服务深度合成管理规定（第六、十六、十七条，2023 年 1 月 10 日施行）. https://www.gov.cn/zhengce/zhengceku/2022-12/12/content_5731431.htm；国家互联网信息办公室、工业和信息化部、公安部、国家广播电视总局 (2025). 人工智能生成合成内容标识办法（第十条，2025 年 9 月 1 日施行）. https://www.cac.gov.cn/2025-03/14/c_1743654684782215.htm；公安部刑侦局、央视新闻 (2026). 警方查获 5 万多条合成动态人脸视频 谁在「复制」你的脸？http://gat.fujian.gov.cn/ztzl/fjjffpzxrx/spjq/202603/t20260330_7117146.htm（福建省公安厅转载）

Always use an old number saved in your contacts when calling back — never dial the number shown on the incoming call or any new number the scammer provides. If the caller cites “poor signal” or “a meeting” to avoid a live video or motion test, assume it is a scam. Motion verification is only a supplementary tool; since AI capabilities keep improving, the only truly reliable step remains cross-channel confirmation. If money has already been transferred, refer to section 2 of this chapter and call 110 immediately to request a stop on the payment. For information on protecting your own facial and voice data, see section 14. This item is rated B because official sources provide identification methods and case examples, but no statistics are given on how much these measures actually reduce fraud rates. The main Ministry of Public Security website (mps.gov.cn) remains inaccessible due to a 521 error; the quoted texts from the Cybersecurity Bureau and Criminal Investigation Bureau were taken from a republishing page run by the Fujian Provincial Public Security Department.