Content Regulation and AI: How China Is Managing the Risks of AI-Generated Media

What do we do when AI can fabricate news, create deepfakes, or simulate a fully functioning human influencer with stunning realism? For China, this is not a futuristic dilemma, but an immediate political problem to be tackled.

As the capabilities of generative AI expand to include the creation of everything from fictitious news articles to virtual celebrities, countries are struggling to control this 'digital wildfire'. In contrast to the rest of the world, China has a well-defined and strict approach to the management of AI-generated content.

China's efforts involve a combination of content restrictions, real-name registration policies, and AI generated content legislation (AIGC). All of these attempts seem to pertain to China’s balancing act between fostering creativity and maintaining ironclad control. This blog analyzes how the Chinese economy is responding to the concerns surrounding AI technologies and what repercussions that poses for businesses and creators, while highlighting why China will continue to be in the spotlight for the global conversation.

______________________________________________________________

The Reason Why Regulation on Computer Generated Media Is Needed.

 

In recent times, the proliferation of chatbots like ChatGPT and deepfake video generators have artificial intelligence technology that mimics creativity, human faces, and speech rising greatly. Such technology brings opportunities in automation, education, entertainment, and marketing; however, it poses immense risks which include: 

• Fake news and misinformation

• Impersonation and fraud

• Political propaganda

• Copyright infringement

• Digital identity manipulation

With over 1 billion people using the internet and a strong government-controlled media system, China does not see AI-generated content (AIGC) merely as a tool. Instead, in their eyes, it poses great national security risk if not monitored or controlled.

 

______________________________________________________________

China Policy Focus: The Deep Synthesis Regulation - 2023

 

In January 2023, China launched one of the first formal policies regulating deep synthesis technology, which includes: 

• Synthetic video, audio, images, and text

• Virtual avatars and digital influencers

• Generative AI systems

 

Main Rules:

1. Content Labeling

All content generated by AI needs to have notifications or markings that show it was made by AI. This applies to synthetic social media profiles, deepfake videos, and virtual news anchors.

2. Real-Name Verification

Real identity of users who intend to develop or use AIGC tools should register. This simplifies the process of tracking misuse and fulfilling responsible governance.

3. Responsibility of the Platform

Baidus, Alibabas, and Tencent’s tech companies are expected to oversee and manage content generated through the AI technology. They are responsible for the moderation of illegal or harmful content and must keep logs of all actions taken for further review.

4. Nothing Shall Be Hidden Behind the Truth

Content created, even if harmless, impersonating someone, tampering with facts, or “jeopardizing national safety” is thoroughly prohibited.

This regulation is the centerpiece in an attempt to “wage the AI arms race while… under the banner of socialist values” an expression cited in numerous state-controlled publications.

________________________________________

Illustration: Supervision of Fake News and Entertainment Reports in Information

AI technology-generated content is not an alien concept to China. The country is unrivaled when it comes to showcase case use—with virtual news presenters at the Xinhau News agency and AI stars on Douyin (the Chinese version of TikTok).

Pilot project: Xin Xiaowei, the virtual news anchor

Xinhua put out Xin Xiao we, the first 24-hour, self-sufficient, female streamlined AI-synthesized voice  and deep synthesis visuals news anchor. It is worth noting that her newscasts have signage stating, “This is a digitally synthesized figure created with artificial intelligence tools.”

Why use the disclaimer? Confusion could arise, regulations could be broken, public trust might be lost in the digital ecosystem. So, it’s best to mitigate things before they happen.

________________________________________

The Role of Big Tech: Baidu, Tencent, Alibaba

AIGC regulation won't be effective without strong enforcement by Chinese technology companies:

• Baidu has already implemented censorship filters on politically sensitive outputs in its ERNIE Bot, a competitor to ChatGPT.

• All developers using Tencent's AI services are required to submit their use cases for review prior to public release.

• Alibaba Cloud provides watermarking, as well as API-enabled content labeling, for generative AI’s creative enterprises.

Their compliance efforts go beyond merely adhering to regulations; they are engineers of self-sufficient ecosystems bound by rules and regulations. This is part of a larger change where compliance is increasingly integrated into system design rather than tacked on afterwards.

________________________________________

Regulating AI Art and Music: Protecting Copyright in the Age of Machines

Protecting intellectual property is one of the most concerning issues with generative technology, particularly because of its implications for creativity.

The following points were provided by China’s National Copyright Administration:

• A work produced by AI can be protected if there is sufficient human involvement in providing creative direction.

• However, models which are trained on data that is copyrighted without permission will infringe intellectual property rights.

Restricting copyright like this poses a problem for AI systems that allow users to generate works “in the style of” well-known Chinese painters or music resembling copyrighted songs.

Example: AI Virtual Idol ‘Ayayi’

Ranmai Technology developed ‘Ayayi’, a virtual influencer who promotes major brands like Gucci and Tesla. She is featured in campaigns for these schools. However, every piece of content bearing her name goes through moderation, and any music/video made using external data must have appropriate licensing documents.

________________________________________

Oversight vs. Innovation Struggle

The balance challenge everywhere including China is how to permit AI advancements while controlling for social stability or veracity. 

Some highlights of China’s approach include: 

• Regulation before an advance rather than punitive after

• Submission delinquency protocols

• Citizen advocacy programs designed to educate the public.

Critics say this method is an infringement on free speech, but proponents see a much-needed boundary in the age of technology.

________________________________________

Global Impact: China's Example for the World

This system of AI content regulation in China is scrutinized by other states, particularly those worried about 

• Elections being compromised and deepfake technology being deployed

• Online hate speech and fake personas

• Hyper-partisan automated lies during war breakouts

In fact, the EU’s AI Act and discussions on AI in the US are adopting elements from China’s approach regarding content labeling and platform responsibility

________________________________________

Final Thoughts: Shaping AI to Build a Media World We Can Trust

Reform is not an option when it comes to what we hear, see, or read online. AI as a technology has to be managed; it is a priority. Enforcement of real name systems and platform responsibility is an attempt by China to encourage trust, truth, and social balance while managing the swift power of technology. 

Labeling standards also governs AI media efficiency, creativity, and scale. Guarding against everything that would undermine trust in AI technology requires some stringent measures. 

Undoubtedly, AI content moderation is the new future of digital governance, and at the center of this is China.