AI-generated short dramas spark copyright and portrait rights violations in China
AI-generated short dramas in China are reportedly causing widespread copyright and portrait rights violations through face-swapping, voice cloning, and unauthorized content remixing. Regulators including the National Radio and Television Administration have launched cleanup operations, and experts are calling for industry-wide governance reforms to address identification challenges and enforcement gaps.
What happened
AI-generated short drama content has rapidly expanded in China, with platforms like Douyin hosting videos carrying an 'film and television editing' label that have accumulated over one trillion views. The technology enables creators to rapidly generate short drama content at low production cost, but has also created massive copyright and portrait rights violations.
According to experts, AI portrait rights violations occur in four main forms: complete AI generation; targeted AI generation using keywords to create images highly similar to celebrities; direct AI face-swapping replacing actors with celebrity faces; and 'stitched' or AI-fused faces combining features from multiple people. Copyright violations include unauthorized reproduction and adaptation of existing films and television works, including direct copying, automated clip extraction, and remixing.
The National Radio and Television Administration launched a month-long 'AI sorcery modification' video cleanup operation in January 2026, focusing on illegal 'AI modifications' of Four Great Classical Novels adaptations, historical dramas, revolutionary films, and model hero productions. The operation removed approximately 23,000 related violation videos and disciplined over 100 violating accounts.
Reported impact
Actresses including Zhao Luosi reported having their faces used without authorization in AI-generated short dramas. Copyright holders report numerous infringements characterized by high volume, high frequency, concealed nature, and rapid spread that harm rights holders' legitimate interests and pose challenges to platform governance and healthy industry development.
The rapid growth of infringement disputes has placed massive pressure on regulation, judicial practice, and platform governance. National court data shows that in 2024, courts adjudicated 125,202 first-instance cases involving copyright network transmission rights infringement, a year-on-year increase of 15.71%, and 5,450 copyright ownership disputes, up 24.94%.
Index analysis
Experts identified three major barriers to identifying and addressing AI portrait violations. First, the lack of uniform 'recognizability' standards: AI-synthesized faces undergo feature modification and detail adjustment, creating a 'similar but not identical' effect that makes it difficult to apply traditional infringement standards, and the industry currently lacks unified clear audit standards. Second, mixed faces and dynamic imagery increase audit difficulty: combined synthetic images blur multiple people's features without clear original prototypes, and dynamic facial features change in real time across video shots, increasing identification difficulty. Third, insufficient technical and data support: the industry lacks mature dynamic facial recognition technology and cannot establish portrait databases covering ordinary citizens, making comprehensive effective monitoring difficult.
Sound infringement presents similar identification challenges. Synthesized voice infringement is harder to identify than visual deepfakes because voice itself is more abstract and depends on subjective perception. For voice actors, performing voice often differs from natural speech. Judging whether synthesized sound constitutes infringement requires comprehensive consideration of tone, pitch, and delivery style across multiple dimensions.
Traditional 'notice-and-takedown' procedures often fail to function smoothly when platforms lack legal authority to make independent rulings on the basis of screenshot comparisons claiming 'high similarity.' The industry urgently needs to construct more refined, operable evidentiary standards.
Governance controls
Experts called for responsibility to be placed on production companies and treatment nodes to be moved forward. Current governance focuses on content review and post-hoc disposal, failing to suppress infringement at the source. They recommended compliance requirements be integrated throughout the creation process centered on production companies, rather than relying solely on platforms' passive post-production review, achieving an upgrade from post-hoc identification to full-process regulation.
Platforms should establish categorized audit mechanisms, build feature databases for high-frequency infringement scenarios like public figures' portraits, use technical means for active early warning and interception, optimize notice-and-takedown procedures, establish convenient complaint channels, and simplify the claims process. Platforms should also improve punishment mechanisms for infringement accounts, taking measures such as traffic limits and bans for malicious infringement and repeat violations.
The National Copyright Office and related departments launched the 'Sword Network 2026' anti-network infringement and piracy operation, focusing on audiovisual works, online literature, and network transmission. Industry associations should lead efforts to develop group standards or industry guidelines covering content identification, rights preservation, infringement comparison, and platform duty of care, ensuring 'all links have standards and all standards are implementable.' Industry associations should establish full-chain credit files for production companies, platforms, and technical service providers.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.