A few days ago, the General Office of the State Council issued the State Council 2023 Legislative Work Plan. Among them, it is worth noting that the draft artificial intelligence law and other preparations are to be submitted to the Standing Committee of the National People's Congress for deliberation, which shows that the mainland's legislation in the field of artificial intelligence will reach a new level.
↑Profile image Illustration IC photo
In fact, the "New Generation of Artificial Intelligence Development Plan" issued by the State Council in 2017 put forward the strategic goal of "three steps", and in 2020, artificial intelligence ethics norms and policies and regulations in some fields were initially established. In 2025, establish a system of AI laws, regulations, ethics and policies, and form AI safety assessment and control capabilities. By 2030, a more complete system of artificial intelligence laws, regulations, ethics and policies will be built.
Yang Yanchao, researcher at the Institute of Law of the Chinese Academy of Social Sciences, director of the Science and Technology and Law Research Center and doctoral supervisor, said in an interview with Red Star News that it is necessary to consider both artificial intelligence risks and industrial development. Artificial intelligence is still in the early stage of technological development, and it is not appropriate to stipulate many issues too carefully or even too strictly. The law should also focus on encouraging industrial development, and should determine the legislative idea of large red lines and negative lists. The so-called big red line and negative list is to draw a red line considering the several major harms that artificial intelligence may bring to social development. For example, artificial intelligence cannot be used for fraud, cannot be used for data attacks, etc.
Why are the conditions ripe for AI legislation?
In the past few years, the establishment of rules and regulations in the field of artificial intelligence in mainland China has begun. In the Provisions on the Administration of Algorithm Recommendations for Internet Information Services (hereinafter referred to as the "Provisions on the Administration of Algorithm Recommendations"), which came into effect on March 1, 2022, the generation of synthetic algorithms is listed as one of the five categories of algorithm recommendation technologies (generation synthesis, personalized push, sorting and selection, retrieval and filtering, and scheduling decision-making), which is subject to supervision.
In the Provisions on the Administration of Deep Synthesis of Internet Information Services (hereinafter referred to as the "Provisions on the Administration of Deep Synthesis"), which came into effect on January 1, 2023, deep synthesis is regarded as the first legislative supervision of algorithms with medium and high risks in the types of algorithm services, further strengthening the supervision of this field.
In order to further promote the regulation and development of the field of artificial intelligence, the Cyberspace Administration of China issued the Measures for the Administration of Generative Artificial Intelligence Services (Draft for Comments) on April 11, 2023.
Based on the above three documents, with the development of algorithms and artificial intelligence technology, the mainland has also formed an iterative path of "algorithm governance - deep synthetic governance - generative artificial intelligence governance" for its governance.
↑ Profile picture Illustration figure worm creative
Professor Yang Yanchao, researcher of the Institute of Law of the Chinese Academy of Social Sciences, director of the Research Center for Science, Technology and Law, and doctoral supervisor, has long studied AI legislation. In Yang Yanchao's view, the introduction of the above management measures reflects that with the advancement of artificial intelligence technology, legislative work has also progressed.
Yang Yanchao explained that the rules related to the recommendation algorithm introduced at the beginning were because many platforms used algorithms to improve services, using user portraits to calculate user preferences, and users did not need to retrieve, and the platform would automatically recommend relevant content. After the recommendation algorithm was widely used, the first thing to introduce was the management rules for the recommendation algorithm. Later, deep synthesis technology appeared, including "face swapping" technology, which can generate speech and images. This process has also triggered people's thinking about rights, including whether personal rights will be violated, so deeply synthetic management rules have been introduced. At the beginning of this year, ChatGPT this generative artificial intelligence caused a series of challenges, and at present, various fields are using large models to complete generative artificial intelligence, generating text, images, and even code, video, you can see that artificial intelligence has a certain ability to integrate data, reasoning, and logical analysis, which is the ability of talents in the past, so it will produce a series of problems such as personal privacy, data security, and intellectual property protection. Concerned about these issues, the rules for the management of generative artificial intelligence have begun to be developed. This is a vein of technological and legal evolution.
The reason why the draft artificial intelligence law was proposed to be included in the legislative plan this year, Yang Yanchao said, is because artificial intelligence technology has grown explosively this year after years of accumulation. In the entire technology industry, the large model has become a breakthrough, and before that, in some local areas, the small model solved certain problems, but was in the technical bottleneck. This time the development of large models has brought completely new capabilities that can be realized. OpenAI's experiments have shown people that AI has produced human-like abilities in the era of big models. In such a special period, while thinking about the development of the industry, legislators and managers will also consider the corresponding risks and ensure the healthy development of the industry by formulating rules.
Yang Yanchao said that the laws that have been formulated in the past in the mainland, including the Personal Information Protection Law and the Data Security Law, as well as the above-mentioned departmental regulations on recommendation algorithms, deep synthesis, and relevant provisions on intelligent driving, have become the early accumulation of artificial intelligence legislation.
Zhang Xin, executive director of the Digital Economy and Legal Innovation Research Center of the University of International Business and Economics, said in an interview with Red Star News that although the above legal norms have a sequential time, they are not a linear development process, and the draft artificial intelligence law has actually been prepared for a long time, and has been promoting, and the legislation of such small cuts at the level of departmental regulations has been advancing rapidly.
Zhang Xin also pointed out that at present, the conditions for artificial intelligence legislation in all aspects are ripe, and there are already some low-level legislation at the next level, which has a certain normative basis for higher-level legislation. On the other hand, the public's digital literacy and awareness of legislative needs, including the compliance awareness of technology companies, as well as relevant technical standards and ethics, have been improved. At the level of the entire institutional foundation, the basis for AI law legislation is relatively mature.
What are the legislative difficulties?
Artificial intelligence technology is still in the early stages of development and is full of challenges
Yang Yanchao believes that the difficulty of the current legislation is that people cannot be said to have a good grasp of the laws of artificial intelligence technology, and they are still in the exploration period. For what kind of risks artificial intelligence technology may produce, it is necessary to summarize and explore, and even some destructive cases can be gradually improved after appearing. Therefore, the current legislation can only be relatively principled legislation.
"The matching between various laws and regulations is a big challenge." For example, at the legal level, data platforms and algorithms are inseparable from artificial intelligence, Zhang Xin said. There have been personal information protection laws, e-commerce laws, cybersecurity laws, data security laws, etc., and what is the relationship between these laws and the upcoming artificial intelligence law, and how it is matched, needs further consideration. The second challenge is that at present, the mainland already has some legislative basis at the level of second-level legislation, such as departmental rules and administrative regulations, and how to match with these existing legal norms below when making higher-level basic laws.
↑Profile image Illustration IC photo
Zhang Xin said that the third challenge is that the field of artificial intelligence is changing very quickly, especially the development of technology, the emergence of generative artificial intelligence. For example, the European Union's Artificial Intelligence Act, if ChatGPT had not appeared when it was formulated, it might not have been such a rapidly changing face today, and it may have basically passed. But because of the advent of ChatGPT, the rhythm it set was disrupted. From the formulation of the EU's "Artificial Intelligence Act", it can be seen that the lag of legislation in the field of artificial intelligence is becoming more and more obvious. Therefore, whether for the mainland or other countries, artificial intelligence legislation is not only inclusive and prudent, but also leaves sufficient flexible space for scientific and technological innovation, and it is very difficult to be forward-looking. This may require an open and dynamic flexible space in legislative technology, including the legislative system.
What does legislation need to focus on?
Artificial intelligence involves all aspects, and it is necessary to pay attention to computing power, algorithms, and data
Artificial intelligence will involve all aspects of production and life, such as content governance, data protection, intellectual property rights, intelligent driving, etc., as well as technical ethics in the development of artificial intelligence technology, and possible infringement issues in artificial intelligence applications and services. Will the law also be as large as a code? What will be necessary in the draft AI law?
↑ Profile picture Illustration figure worm creative
Yang Yanchao said that artificial intelligence has performance in various fields, such as unmanned driving, intellectual property, finance, etc., and the legal relationships, rights and obligations, and legal responsibilities shown in each field are different. The legislative process is likely to be gradually sorted out and summarized from the problems arising from the sectoral laws.
Yang Yanchao for example, for example, in the field of unmanned driving, the legal rules produced have been very complex, but its characteristics are different from generative artificial intelligence, what needs to be done is to change traditional traffic laws, in the past drivers have become machines, which will fundamentally change the understanding of traffic accident fault liability and tort liability. The same is true in other fields, legislation should be deeply cultivated in subdivided areas, and then abstracted to do codification and upgrading.
Zhang Xin also believes that reference is made to the Personal Information Protection Law and the Cybersecurity Law. As a foundational law, AI law may not be as long as the EU's Artificial Intelligence Act; It may be that the rights and responsibilities of artificial intelligence users, service providers, developers, and deployers of these subjects are clear, and will not be as large as a code.
So what should AI legislation have? Yang Yanchao said that no matter what field of artificial intelligence legislation, it is inseparable from the provisions of the three elements of artificial intelligence and the consideration of three principles. The so-called three elements are: computing power, algorithms, and data, because no matter what kind of artificial intelligence problems are solved, these three elements are inseparable. The so-called three principles are: industrial development, rights protection, ethical considerations, and the laws formulated to protect people's rights must withstand the consideration of modern people's ethics, and can also promote the development of industry.
How to deal with the relationship between risk and development?
Experts recommend drawing a red line: AI cannot be used for scams, data attacks
How can laws be enacted to avoid chaos without restricting technological development and achieving good law and good governance? Yang Yanchao said that this requires considering both artificial intelligence risks and industrial development. Because it is still in the early stage of technological development, it is not appropriate to stipulate many issues too carefully or even too strictly. The entire law should also focus on encouraging industrial development, and should determine the legislative ideology of large red lines and negative lists. The so-called big red line and negative list is to draw a red line considering the several major harms that artificial intelligence may bring to social development. For example, artificial intelligence cannot be used for fraud, data attacks, etc. Draw a red line, and no industrial development can cross it. This not only gives great space for industrial development, but also ensures that the basic security of human society will not be violated.
Zhang Xin said that AI governance should carry out agile governance on the one hand, and at the same time ensure that governance is resilient. After proposing this governance structure and goals, it is necessary to rely on a series of governance tools to implement it, such as in the field of artificial intelligence, algorithm impact assessment, algorithm certification, algorithm audit, etc. are all very good governance tools. With these tools, it can be guaranteed that it is constantly homomorphic with the development of these technologies, which is a continuous dynamic adjustment and tuning process. In the field of artificial intelligence, whether it is legislation or the entire governance process, it is necessary to rely on a series of governance tools to implement it.
Red Star News chief reporter Wu Yang Intern Li Chenxiao reported from Beijing
Edited by Zhang Li Edited by Feng Lingling
(Download Red Star News, there is a prize!) ）