In September 2025 a court in Shangyu, Zhejiang, delivered a first-instance judgment in a case involving AI face-swap technology used to buy Maotai liquor. According to public reports, two defendants used face-swap technology to bypass platform identity verification and buy in bulk, and were convicted of illegally controlling computer information systems, sentenced to fixed-term imprisonment of one year one month to one year four months, suspended, with fines. A third defendant, involved with personal information, was convicted of infringing citizens’ personal information, sentenced to eight months, suspended for one year, with a fine. None appealed.
One act, two offences
The most notable feature is that a single misuse of technology triggered two different offences.
Illegally controlling computer information systems targets interference with normal system operation. Bypassing identity verification through a face swap is, in substance, unauthorised intervention in the platform’s authentication logic, causing it to operate in an unintended way.
Infringing citizens’ personal information targets unlawful acquisition and use of personal information. Face swapping usually needs another person’s facial information as source material, which falls directly within personal information protection.
AI-related criminal risk is rarely singular. Facial information is both an identity credential and personal information; bypassing verification is both interference with a system and potential information misuse. See our note on criminal compliance for AI companies.
Three risk lines in face-swap misuse
Identity verification. Any context using a face as a credential — payment, government services, platform verification, access control — is a high-incidence area. See our note on face recognition filing.
Personal information. Facial data is sensitive personal information requiring separate consent and a lawful basis. Training a face-swap model on another person’s face is high risk. See our note on obligations of major personal information processors.
Content labelling. Swapped videos are synthetic content, and labelling duties cannot be avoided. See our note on the AI content labelling measures.
What it means for companies and developers
For companies offering face-swap or digital-human capabilities, the lesson is that technology may be neutral, but capability delivery needs risk controls.
First, set allow-lists and deny-lists by use case, with stricter review for calls touching identity verification or financial payment.
Second, prohibit specified uses in user terms, backed by anomaly detection and handling. See our note on AI commercial contract clauses.
Third, retain logs of technical calls, so the company can show it exercised reasonable care in a dispute.
For ordinary users, the case is also a reminder: using “technical means to improve purchase success” to circumvent platform rules can carry consequences far beyond the expected gain.
This article is based on publicly reported judgment information. It is for general reference only and does not constitute legal advice.