How AI Deepfakes Threaten Privacy and What Law Is Doing About It

On May 28, 2020, the 13th National People's Congress passed China's Civil Code, which took effect on January 1, 2021. Article 1019 explicitly prohibits the use of technical means, such as AI face‑swapping, to infringe on portrait rights, marking the first legal focus on deepfake technology.

Any individual or organization must not, without consent, create, use, or publicly display a person's portrait by distorting, defaming, or employing information‑technology methods such as forgery.

Deepfake, a term coined in December 2017, refers to AI‑generated media that swaps faces or alters expressions, often producing realistic yet false videos and images. While the technology originated from deep learning research in image classification, its accessibility has dramatically increased, lowering the barrier for creating convincing fake media.

Deepfake Proliferation

Tools like DeepFaceLab have gained popularity on GitHub, boasting thousands of stars, and applications such as ZAO allowed users to replace faces in short videos within seconds, quickly topping app store charts before being shut down due to privacy concerns.

Deepfake content is frequently used for entertainment, but malicious uses—especially in political or pornographic contexts—have sparked legal and societal alarm. Reports indicate a rapid growth of deepfake videos, with the majority being pornographic and amassing billions of views.

Several high‑profile incidents, such as fabricated videos of politicians, have demonstrated the potential for deepfakes to influence public opinion and even incite unrest.

Mitigating Harm with Technology, Ethics, and Law

Researchers and companies are developing detection methods. In 2019, DARPA launched the SemaFor project to automatically detect forged media. Facebook, Microsoft, Google, Amazon, and academic institutions organized a deepfake detection challenge, providing large datasets and funding.

Platforms now label or remove deepfake content deemed harmful. Facebook warns users about AI‑generated misinformation before elections, while Twitter marks synthetic media and may delete content that threatens safety or public order.

According to the Tort Liability Law, network service providers must delete, block, or disconnect infringing content; failure to act makes them jointly liable with the infringing user.

Legal experts emphasize that technical detection must be complemented by clear regulations. Recent Chinese regulations, such as the "Network Audio‑Video Information Service Management Measures," prohibit the use of deep learning and related technologies for illegal activities, reinforcing the need for both ethical education and statutory safeguards.

Educators like Professor Chu Ying advocate for mandatory computer ethics courses to raise moral standards, but stress that legislation provides the enforceable backbone needed to prevent technology misuse.