China to review generative AI before approving its use

China has announced that it will carefully review generative AIs (such as ChatGPT) before authorizing their use in the country. The Cyberspace Administration of China (CAC) has published the Generative AI Service Governance Measures, a series of guidelines that companies seeking to offer products with the technology in the country must comply with. While the document is only a draft, it leaves clarity or intent to establish government regulation.

CAC states that the measures were developed in accordance with the Data Security Act, the Personal Information Protection Act and the Internet Safety Act. Under this scheme, the government specified that generative AI should reflect the values of socialism, avoid discrimination, respect copyright, and provide truthful and accurate content.

The government's announcement comes at a key moment for this technology in the Asian country. Companies such as Baidu and Alibaba have already submitted their generative AI, which they plan to deal with in ChatGPT. While they are not yet at the GPT-4 level, Chinese alternatives are aiming to offer similar options for those using each company's services.

Here are the rules that generative AI must comply with in China

The rules will apply to research, development, and generative artificial intelligence products capable of creating text, images, videos, sounds, and other content based on algorithms and models. While the idea of regulation represents an obstacle to the development of AI, the CAC has made it clear that the state supports innovation.

The provision of generative artificial intelligence products or services must comply with the requirements of laws and regulations, respect public morals, public order and morality, and comply with the following requirements:

1. The content generated by generative artificial intelligence should reflect the fundamental values of socialism and should not contain the overthrow of state power, the overthrow of the socialist system, incitement to divide the country, undermine national unity, promote terrorism, extremism and incite hatred. and ethnic discrimination, violence, misinformation, etc. obscene and pornographic, false information and content that can disrupt the economic and social order.
   In the process of developing algorithms, selecting training data, generating and optimizing models, and providing services, measures will be taken to prevent discrimination based on race, ethnicity, religion, country, region, gender, age, profession, etc.
   Intellectual property rights and business ethics must be respected and no algorithms, data, platforms or other advantages should be used to promote unfair competition.
   Content generated by generative artificial intelligence must be truthful and accurate, and measures will be taken to prevent the generation of false information.
   It is necessary to respect the legitimate interests of others, prevent damage to the physical and mental health of others, damage image rights, reputational rights and personal privacy, and infringe intellectual property rights. Illegal acquisition, disclosure and use of personal information, confidentiality and trade secrets is prohibited.

An important point of this regulation is that it makes companies or individuals responsible for content created using artificial intelligence. Service providers (such as Baidu or Alibaba) will need to ensure that the data they use to train their AI is legal. The database must comply with applicable laws and must not contain personal information, infringing content or any other laws.

The China Cyberspace Administration has said that products will undergo a security assessment before offering generative AI services. In the event that they are found to be used to send spam or negative comments to the government, the service will be immediately suspended.

Once approved, these measures will apply to all AI-based products and services in China.