In late November 2023, a citizen reported to the Yizheng City Public Security Bureau that a relative had used AI-generated dynamic facial videos to bypass facial recognition systems and steal QQ gaming accounts. Law enforcement launched an investigation the following day, subsequently apprehending intermediaries across Shanghai, Guangdong, and Yunnan, and later identifying and arresting the core technical team members in Shanxi province.
The investigation revealed that starting in July 2020, a suspect recruited two relatives to serve as technical personnel, utilizing software to mass-produce dynamic facial videos for cracking facial recognition systems on QQ, WeChat, and multiple market supervision platforms. Exploiting the biometric verification features of these platforms, the group accepted business including WeChat account unlocking, QQ phone number rebinding, and enterprise WeChat registration. Intermediaries advertised their illegal services online, charging clients and subcontracting the work to the technical team, earning margins of 20 to 200 yuan per transaction.
The criminal operation also involved the long-term acquisition of ID card data containing facial photographs, which were used to generate realistic dynamic facial videos. These AI-produced videos were deployed across various types of criminal activity. From 2020 through December 2023, one defendant used the AI-generated videos to bypass identity verification systems, rebinding phone numbers on QQ accounts and subsequently stealing virtual items for resale. Another individual began actively participating in similar crimes in late 2023.
Beyond gaming account theft, the dynamic facial videos were used to falsify business registration modifications. Former company registration agents, encountering obstacles when handling corporate data changes due to missing or deceased legal representatives, discovered they could use the AI service to circumvent verification processes. From 2022 onward, these agents illegally provided clients with unauthorized modification services to commercial registration systems in exchange for fees.
Technical Method Revealed
The criminal groups stored AI-generated facial videos on computers loaded with software designed to tamper with or replace system camera feeds. When facial recognition verification was required, the software would substitute the real-time camera view with the pre-recorded synthetic video, thereby bypassing the identity verification mechanism and completing authentication unnoticed.
During 2025, prosecutors conducted forensic experiments using the same AI software to reconstruct the entire process of creating and uploading dynamic facial videos. The technical analysis demonstrated that the suspects illegally controlled computer information systems by altering platform operations to accept forged facial videos, without directly modifying or deleting system functions. This established the legal basis for charging them with illegal control of computer information systems.
The prosecution determined that the AI facial recognition bypass was not an isolated crime, but was interconnected with other offenses including the leakage of citizens' personal information, theft of network accounts, fraudulent commercial registration, and telecommunications fraud. A comprehensive supplementary investigation was ordered, leading to the arrest of 74 individuals involved in illegally obtaining data, infringing on citizens' personal privacy, and committing fraud. Total criminal proceeds amounted to 7.23 million yuan.
Financial Gains and Sentences
The main technical team members accumulated illegal profits exceeding 5.47 million yuan, with one individual earning 3.28 million yuan and the two other technical members receiving 1.09 million yuan each. Intermediaries earned over 30,000 yuan, while those involved in account theft profited nearly 270,000 yuan. By November 2025, prosecutors had indicted 36 individuals on charges including illegal control of computer information systems, illegal acquisition of computer data, facilitating criminal activities through information networks, and violating citizens' personal information protection laws. Seven others received non-prosecution decisions due to minor involvement.
From December 2025 to March, courts sentenced the 36 defendants to prison terms ranging from six months detention to three years imprisonment, with some receiving suspended sentences and all subject to fines. All illegal gains were confiscated.
In response to the exposed vulnerabilities in facial verification for online commercial registration, prosecutors have transferred case files to relevant provincial market regulation authorities, urging them to strengthen dynamic facial verification mechanisms. To date, more than 400 instances of illegal commercial registration changes using AI-generated facial videos have been identified for corrective action.