On September 26, the National New Generation Artificial Intelligence Governance Professional Committee released the "New Generation Artificial Intelligence Ethical Specifications" (hereinafter referred to as "Normative"), aiming to integrate ethics into the entire life cycle of artificial intelligence, and to engage in artificial intelligence related to Natural persons, legal persons and other related institutions in the activities provide ethical guidance.

As early as January 5 this year, the National Information Security Standardization Technical Committee issued the "Practice Guidelines for Cyber ​​Security Standards - Guidelines for Prevention of Ethical Security Risks in Artificial Intelligence" (hereinafter referred to as the "Guidelines"), indicating that relevant organizations or individuals are carrying out artificial intelligence. When researching and developing, designing and manufacturing, deploying and applying related activities, we should fully identify, prevent and control the ethical security risks of artificial intelligence.

Compared with the "Guidelines", the "Specifications" issued this time have made a more detailed explanation of privacy protection and data security, and also paid attention to technical ethical issues such as algorithm bias.

Pay attention to data and privacy security, emphasize algorithmic ethics

The "Normative" puts forward six basic ethical requirements, including improving human welfare, promoting fairness and justice, protecting privacy and security, ensuring controllability and credibility, strengthening responsibility, and improving ethical literacy. Among them, data and privacy security content runs through the specific ethical requirements of specific activities such as artificial intelligence management, research and development, and supply.

The "Specifications" point out that all kinds of artificial intelligence activities should fully respect the rights of personal information such as knowledge and consent, and process personal information in accordance with the principles of lawfulness, legitimateness, necessity and integrity to protect personal privacy and data security, and shall not damage individual legal data rights and interests, and shall not use them. Illegal collection and utilization of personal information by stealing, tampering, leaking, etc., and shall not infringe upon personal privacy rights.

In terms of management, we must fully respect and protect the privacy, freedom, dignity, security and other rights and other legitimate rights and interests of relevant subjects, and prohibit improper exercise of power to infringe on the legitimate rights and interests of natural persons, legal persons and other organizations.

In terms of R&D, we must improve data quality, and strictly abide by data-related laws, standards and specifications in the data collection, storage, use, processing, transmission, provision, and disclosure, and improve data integrity, timeliness, consistency and specifications in the process of data collection, storage, use, processing, transmission, provision, and disclosure, and improve data integrity, timeliness, consistency and specifications. and accuracy.

In terms of supply, we must strengthen the quality monitoring and use evaluation of artificial intelligence products and services, avoid personal safety, property safety, user privacy, etc. caused by design and product defects, and shall not operate, sell or provide non-compliant quality standards. Products and services.

Algorithm ethics is also one of the key points of this specification. The "Specifications" clearly state in the "R&D Specifications" that in data collection and algorithm development, we must avoid bias and discrimination, strengthen ethical review, fully consider differentiated demands, avoid possible data and algorithm bias, and strive to achieve universalization of artificial intelligence systems Benefit, fair and non-discriminatory. At the same time, we must enhance safety and transparency, improve transparency, interpretability, understandability, reliability and controllability in the algorithm design, implementation and application links, enhance the resilience, adaptability and anti-interference capabilities of artificial intelligence systems, and gradually increase the resilience, adaptability and anti-interference capabilities of artificial intelligence systems. Implement verifiable, auditable, supervisable, traceable, predictable, and trustworthy.

Technical ethics and data security have become the focus of market supervision

In recent years, the issue of ethical governance of artificial intelligence has attracted increasing attention. Data security and algorithmic ethical compliance have always been the focus of attention in my country's artificial intelligence industry, and relevant systems are constantly being improved.

On July 14, 2021, the "Regulations on the Promotion of Artificial Intelligence Industry in Shenzhen Special Economic Zone (Draft)" drafted by the Organization Department of the Standing Committee of the Shenzhen Municipal People's Congress requires artificial intelligence enterprises to "establish ethical risk positions" and "perform ethical review and risk assessment responsibilities", and It is clearly stipulated that a series of behaviors such as infringement of personal privacy and personal information protection, damaging national security and social public interests, and algorithmic discrimination are prohibited in carrying out artificial intelligence research and application activities.

On July 28, 2021, the Ministry of Science and Technology issued the "Guiding Opinions on Strengthening the Governance of Science and Technology Ethics (Draft for Comments)", which clarified the basic requirements such as ethics first and agile governance, and put forward five scientific and technological ethics principles, requiring "engaging in life." Institutions of scientific and technological activities such as science, medicine, artificial intelligence, etc. whose research content involves sensitive areas of scientific and technological ethics, a scientific and technological ethics (review) committee should be established."

In addition, the "Data Security Law of the People's Republic of China" and the "Personal Information Protection Law of the People's Republic of China" were successively issued in June and August this year, and the importance of data and information security has been emphasized like never before. Under the policy trend, AI companies are paying more and more attention to the compliance issues of data security and algorithmic ethics when financing and listing.

In the past year, "AI concept stocks" such as Yitu, Megvii, Yuncong, Haitian Aisheng, and Yuntian Lifei have flocked to IPO, and data compliance and technological ethics have been the focus of review. In May this year, during the IPO of Megvii Technology, the Shanghai Stock Exchange conducted an inquiry on technology ethics for the first time, requiring Megvii to disclose the company's organizational structure, core principles, internal control and implementation in artificial intelligence ethics. Yuncong Technology was once suspended from issuance and listing review due to expired financial information. Yitu Technology has withdrawn its application and its issuance and listing review has now been terminated.

Against the backdrop of increased regulatory efforts, SenseTime, one of the "Four AI Little Dragons", recently submitted a prospectus for listing in the Hong Kong stock market, and set up a special chapter in the prospectus to fully disclose the privacy and AI governance links, indicating that it is in business. How to protect data privacy and personal information, it also said that it will follow three principles of artificial intelligence ethics: sustainable development principle, people-oriented principle and technical controllable principle.

Wang Qiongfei, founding partner of Zhejiang Kenting Law Firm, said in an interview with 21 reporters that for the AI ​​industry and technology companies, in the short term, given that data compliance and AI governance will consume a lot of energy and costs; In the medium term, listed technology companies will take the initiative to disclose personal privacy and AI governance work, which will undoubtedly serve as an example and guide for other companies, thereby reducing data standardization and reducing the workload of data governance; in the long run, this move will also It will accelerate the healthy development of the AI ​​industry and gradually realize the balance between artificial intelligence innovation and supervision while respecting and protecting personal privacy.