When AI can replicate human faces with one click and produce hyper realistic dynamic videos, the age old adage “seeing is believing” is being quietly upended by deep synthesis technologies. AI face synthesis brings creative benefits, but it also poses multiple hidden risks to data security, personal rights and social trust. As technological innovation races ahead, safeguarding compliance boundaries for face synthesis and bolstering the barrier of digital trust has become an imperative for the whole society.
The human face is highly sensitive biometric information that serves as an irreplaceable digital “ID card”. Unlike passwords, which can be reset, facial features, if illegally collected, replicated or circulated, can cause irreversible harm. Recent police investigations across multiple regions reveal that criminals acquire online photos for a few yuan apiece, then leverage AI to generate dynamic facial videos with blinking and nodding gestures. These synthetic videos are used to mass register and resell real name accounts, and even bypass liveness verification to steal funds and launder money. In this black market industrial chain, upstream actors steal citizens’ portraits, mid stream operators deploy AI tools for one click synthesis, and downstream actors exploit these outputs for fraud and money laundering, turning ordinary people’s biometric data into profit making commodities. Meanwhile, infringement disputes keep emerging, where portraits of public figures and internet users are used without permission in short dramas, live stream marketing and malicious parodies. Court rulings draw a clear baseline: if an AI generated figure is recognizable as a specific natural person, unauthorized use still constitutes portrait right infringement, even when marked “AI fictional”. Labelling content as AI generated is no “shield” against illegal conduct.
Technology applications must abide by the rule of law. The Civil Code of the People’s Republic of China explicitly prohibits portrait right infringements perpetrated through information technology forgery. The Provisions on the Administration of Deep Synthesis Internet Information Services mandate real name registration, content review and mandatory labelling for AI generated content, and ban unauthorized use of others’ biometric information such as facial and voice features. The Interim Measures for the Administration of Generative Artificial Intelligence Services further consolidate platform accountability. This systematic legal framework sends a clear message: Any tool shall be applied under the principles of informed consent, respect for personal dignity and risk prevention. As primary technology providers, platforms must not prioritize product usability at the cost of risk control. Risk assessment should be embedded into product research and development. Platforms need to improve identification and traceability capabilities, block loopholes for abuse by cyber criminal groups, and fulfill due obligations for content review and removal of illegal information.
Mitigating risks stemming from AI face synthesis calls for multi party collaboration to build a multi layered protection network. Regulators should carry out regular crackdowns targeting the black market chain of facial data resale and unlawful deep synthesis. Enterprises should optimize identity verification systems and avoid over reliance on facial recognition alone; cross verification should be adopted to reduce vulnerabilities to forged videos. Through cyber security publicity campaigns, the public should be better informed about deep fake identification, getting a “cognitive vaccine” to foster prudent awareness that “videos may not represent reality”.
Trust between people underpins the digital economy. Blurred lines between truth and forgery and eroded public trust would undermine the very foundation of all digital services.
By Vicky Jia, Office Supervisor