# China's Deep Synthesis Regulations

China's Deep Synthesis Regulations, effective January 10, 2023, are administrative rules governing deepfakes and synthetic media, requiring labeling, consent, and content moderation by providers.

China's Deep Synthesis Regulations, formally the "Provisions on the Administration of Deep Synthesis Internet Information Services," took effect on January 10, 2023. Issued by the Cyberspace Administration of China (CAC) along with three other ministries, these rules are among the first national-level legal frameworks specifically targeting deepfakes and synthetic media. They impose binding obligations on organizations and individuals that use deep synthesis technology to generate or alter content, including text, images, audio, and video, within the People's Republic of China.

The regulations define deep synthesis as technology that uses [machine-learning](https://www.wikiprompt.org/wiki/machine-learning) and [deep-learning](https://www.wikiprompt.org/wiki/deep-learning) methods to generate or edit text, images, audio, or video, including techniques for face swapping, voice cloning, and virtual scene creation. The rules apply to both commercial providers and non-commercial users who employ such tools, covering services offered over the internet, including mobile apps, websites, and software platforms. The stated purpose is to safeguard national security, public interests, and the lawful rights of citizens, while promoting the healthy development of [generative-ai](https://www.wikiprompt.org/wiki/generative-ai) technologies.

## Scope and Key Definitions

The regulations explicitly cover a broad range of synthetic media. This includes deepfake videos, realistic voice synthesis, text-to-image generation, and AI-generated virtual anchors. The CAC defines "deep synthesis service providers" as any entity that offers these technologies to the public, either as a primary service or as a feature within a larger platform. Users who employ such services for public dissemination are also subject to certain provisions, particularly around labeling and content legality.

A notable aspect of the rules is their extraterritorial reach. Providers that offer services to users within China, even if based overseas, are expected to comply. The regulations also require that service providers conduct security assessments before launching new products or features that use deep synthesis, and they must file their algorithms with the CAC, a process similar to the existing algorithm recommendation regulations that took effect in March 2022.

## Labeling and Transparency Requirements

One of the central requirements is mandatory labeling. Any content generated or altered using deep synthesis that could mislead the public must be clearly marked as synthetic. The regulations distinguish between "explicit labels" and "implicit labels." Explicit labels are visible to users, such as a watermark or a text notice. Implicit labels are embedded in the file's metadata or digital signature, allowing for machine-readable identification.

For realistic face-swapping or voice-cloning content, the provider must add a prominent notice that the content is AI-generated or edited. The rules also require that providers ensure the labels are not easily removed or tampered with. Additionally, if synthetic content is used in news reporting, the provider must clearly indicate the synthetic nature to avoid confusing it with real footage. This labeling requirement extends to content that is uploaded by users, placing a duty on platforms to detect and label such content when it is published.

## User Consent and Personal Rights

The regulations place strong emphasis on protecting individual rights. Providers must obtain explicit consent from individuals before using their biometric information, such as faces or voices, in deep synthesis. This consent must be informed and voluntary, and users have the right to withdraw it. The rules prohibit the use of deep synthesis to create or distribute content that infringes on a person's reputation, privacy, or other lawful rights.

Furthermore, providers are required to establish mechanisms for users to report violations, such as non-consensual deepfakes. They must also provide a way for affected individuals to request removal of content that uses their likeness without permission. The regulations explicitly forbid the use of deep synthesis to produce fake news, defamatory material, or content that disrupts social order. This aligns with China's broader cybersecurity and data protection laws, including the Personal Information Protection Law (PIPL) effective November 2021.

## Content Moderation and Security Obligations

Service providers must implement robust content moderation systems. They are required to review all synthetic content for illegal or harmful material, including content that promotes terrorism, violence, pornography, or subversion of state power. The regulations mandate that providers use technical means to identify synthetic content and prevent its misuse. This includes maintaining logs of user activities and content generation for at least six months, subject to data retention laws.

Providers must also ensure that their algorithms do not reinforce illegal or unethical behaviors. The rules require that deep synthesis services be designed with safety features, such as preventing the generation of content that impersonates government officials or public figures in a misleading way. In cases where synthetic content could affect public opinion or national security, providers are expected to cooperate with authorities and take immediate action to remove or block such content.

## Algorithm Filing and Security Assessments

A key compliance mechanism is the requirement to file algorithms with the CAC. Providers must submit detailed information about their deep synthesis algorithms, including their purpose, data sources, and potential risks. This filing process is similar to that for recommendation algorithms under the 2022 regulations. The CAC maintains a public registry of filed algorithms, and failure to file can result in penalties.

Security assessments are required before launching new services or major updates. These assessments evaluate the potential risks to national security, social stability, and individual rights. The assessments must be conducted by qualified third-party organizations or internal teams, and the results must be submitted to the CAC. The regulations also require that providers establish emergency response plans for security incidents, such as large-scale deepfake campaigns.

## Enforcement and Penalties

Violations of the regulations can lead to administrative penalties, including fines, suspension of services, and revocation of licenses. The CAC and other regulators have the authority to conduct inspections and demand corrective actions. In severe cases, criminal liability may apply under China's Criminal Law, particularly for offenses related to fraud, defamation, or spreading false information.

Since the regulations took effect, the CAC has actively enforced them. In 2023, several platforms were publicly criticized for failing to label synthetic content adequately. The CAC has also issued guidance on specific use cases, such as AI-generated virtual anchors in livestreaming, requiring that they be clearly identified. The enforcement approach reflects a broader trend in China of regulating [artificial-intelligence](https://www.wikiprompt.org/wiki/artificial-intelligence) through sector-specific rules rather than a single comprehensive law.

## Impact on Industry and Global Context

The regulations have had a significant impact on both domestic and international companies. Major Chinese tech firms, including [alibaba-cloud](https://www.wikiprompt.org/wiki/alibaba-cloud) and other cloud providers, have integrated compliance features into their AI services, such as automatic watermarking and content filtering. International companies offering deep synthesis tools to Chinese users have also adjusted their practices, often by geo-blocking or adding local compliance measures.

Globally, China's approach has been influential. The European Union's AI Act, which entered into force in August 2024, includes similar transparency requirements for deepfakes, though with different enforcement mechanisms. The United States has taken a more fragmented approach, with state-level laws and federal guidelines. China's regulations are notable for their comprehensive scope and the integration of algorithmic filing, which is unique among major economies.

## Future Developments and Challenges

As of 2025, the CAC has signaled that it will update the regulations to address emerging technologies, such as [large-language-model](https://www.wikiprompt.org/wiki/large-language-model)-based content generation and real-time deepfake video. The rapid evolution of [generative-ai](https://www.wikiprompt.org/wiki/generative-ai) poses challenges for enforcement, as detection methods lag behind generation capabilities. The regulations also face criticism from some scholars who argue that the labeling requirements are insufficient to prevent misinformation, while others contend that the rules are overly restrictive and hinder innovation.

Despite these challenges, the regulations have established a baseline for responsible use of deep synthesis in China. They have prompted the development of technical standards for watermarking and content provenance, and they have encouraged research into detection algorithms. The long-term effectiveness of the regulations will depend on ongoing updates and the ability of regulators to keep pace with technological change.

## Comparative Analysis with Other Jurisdictions

Compared to the EU's AI Act, which categorizes deepfakes as a transparency risk, China's regulations are more prescriptive, requiring specific labeling formats and algorithmic filings. The EU relies on general transparency obligations and risk management, while China mandates detailed technical compliance. In the United States, the absence of a federal law has led to a patchwork of state regulations, such as California's law on political deepfakes, which took effect in 2020. China's centralized approach allows for consistent enforcement but raises concerns about overreach and censorship.

The regulations also interact with China's data localization requirements, as providers must store user data within China. This has implications for international companies that operate globally. The requirement for algorithm filing has been compared to similar measures in other sectors, such as the 2021 rules on recommendation algorithms, which have been used to control information flow. Overall, China's deep synthesis regulations represent a pioneering effort to govern synthetic media, with both supporters and critics watching their implementation closely.

---
Source: https://www.wikiprompt.org/wiki/china-deep-synthesis-regulations
License: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/)
Last updated: 2026-09-13T03:51:30.116514+00:00