Features of the development of legal regulation of recommendation technologies in business in Russia and China
In the article, the author examines approaches to the legal regulation of recommendation algorithms in Russia and China, analyzes the relevant concepts, requirements for transparency, control and monitoring within the designated topic. In general, possible problems in practice related to access to information systems are noted due to the need to attract highly qualified specialists, the volume of analyzed data, and the risk of leakage of confidential information.
Keywords: recommendation technologies, recommendation systems, recommendation algorithms, information technologies, algorithmic recommendation technologies, user preferences, entrepreneurial activity.
In the context of active digitalization of various areas of society, entrepreneurial activity as a kind of economic activity is also beginning to acquire the latest features in this area.
V.V. Monastyrev in his scientific article "Recommendation system based on user actions in a social network" notes that when using social networks, online services, users leave various information about themselves, such as comments, photos, geotags, etc. As a result, this information can be used for the operation of special recommendation systems <1>.
--------------------------------
<1> Monastyrev V.V., Drobintsev P.D. A recommendation system based on user actions in a social network // Proceedings of the ISP RAS. 2020. N 3. pp. 100 - 102.
I.Y. Pashchenko notes that the functioning of the recommendation system should take into account not only the positive experience expressed in creating comfortable conditions for information consumption, but also the existing risks to information security of both an individual (user) and the entire state and society as a whole. These risks, for example, can be expressed in the dissemination of false information, distortion of individual perception, impact on the formation of public opinion, lack of transparency due to the complexity of systems <2>.
--------------------------------
<2> Pashchenko I.Y. Prospects of regulatory regulation of recommendations in information systems that are user services // Modern scientific research and innovations. 2022. N 9.
E.V. Talapina notes that the use of recommendation algorithms can also pose risks to fundamental human rights, in particular, equality, the right to privacy and personal data protection may suffer, and discrimination against data processing criteria should not be allowed <3>.
--------------------------------
<3> Talapina E.V. Algorithms and artificial intelligence through the prism of human rights // Journal of Russian Law. 2020. N 10. pp. 25-39.
The development of recommendation algorithms as one of the effective ways to promote a product, work or service, in general, content stimulates legislation to change in new conditions.
Such transformations are primarily due to the idea of observing transparency and legality, which is confirmed by the position of the author of the legislative initiative on the legal regulation of the issue under consideration in the Russian Federation.
State Duma Deputy Anton Vadimovich Gorelkin points out the need to comply with the two above ideas, since the system of personal recommendations is used by virtually every large digital platform, while collecting data not only presented by the person himself, but also obtained as a result of analyzing activity from various devices, up to the use of a neural network processing photos and video files <4>.
--------------------------------
<4> Deputies supported amendments on recommendation algorithms of Internet resources // State Duma of the Russian Federation: official website.
The People's Republic of China (hereinafter referred to as the PRC) is generally ahead of Russia in the legal regulation of the use of recommendation algorithms, as well as artificial intelligence in general, which is reflected in a number of special regulatory legal acts.
1. Administrative Regulations on Recommendation Algorithms for Internet Information Services, which entered into force on March 1, 2022 (hereinafter referred to as the PRC Regulation on Recommendation Algorithms) <5>.
2. Provisions on the management of deep synthesis in the Internet information service, which entered into force on January 10, 2023. <6>
3. Temporary regulations on the management of generative artificial intelligence services, entered into force on August 15, 2023. <7>
--------------------------------
<5> The State Chancellery of the InternetInformation of the People's Republic of China: official website.
<6> State Chancellery of Internet Information of the People's Republic of China: official website.
<7> State Chancellery of Internet Information of the People's Republic of China: official website.
In general, these legal acts impose certain obligations on service providers, technical personnel, users, as well as on commercial organizations, including online platforms.
The ultimate goal of these documents is to eliminate the risks associated with content generated using artificial intelligence, as well as to protect the national and social security of the People's Republic of China.
In Russia, this direction received legal formalization on July 26, 2023 with the adoption of the Federal Law "On Amendments to the Federal Law "On Information, Information Technologies and Information Protection" and the Federal Law "On Communications" <8>, which entered into force on October 1, 2023.
--------------------------------
<8> Federal Law No. 426-FZ of December 2, 2019 "On Amendments to the Law of the Russian Federation "On Mass Media" and the Federal Law "On Information, Information Technologies and Information Protection" // Official Internet Portal of Legal Information.
The PRC Regulation on Recommendation Algorithms in Article 2 formulates a legal approach to defining the concept of "recommendation technology algorithm" - the use of algorithm technologies such as generation and synthesis, personalized approach, sorting and selection, search and filtering, planning and decision-making to provide information to the user.
In Russia, as a result of legislative changes, a legal approach to the concept of "recommendation technologies" has also appeared. At the same time, the amendments made to the Federal Law "On Information, Information Technologies and Information Protection" <9> (hereinafter - the Law on Information in the Russian Federation) do not imply the addition of Article 2 of the said Law with an appropriate concept, but give a number of signs in Articles 10.2-2 of the Law on Information as amended on October 1, 2023.
--------------------------------
Federal Law No. 149-FZ of July 27, 2006 "On Information, information Technologies and information Protection" // Official Internet portal of Legal Information.
Among these features, the following can be distinguished: information technology; activities for the collection, systematization, analysis and provision of information; information related to user preferences.
Combining these features into a single structure, we will get the following legal approach: recommendation technologies are information technologies for providing information based on the collection, systematization and analysis of information related to user preferences.
Comparing the presented definitions, it can be concluded that the PRC's approach is more specific in terms of listing the actions included in the process of the recommendation algorithm. Meanwhile, both definitions have a key qualifying feature, expressed in the procedure for working with personalized data in order to provide information that corresponds to the user's preferences.
Turning to one of the fundamental scientific studies in this field, Recommender Systems: An Overview <10>, two key features can be identified that determine the specifics of recommendation systems in general.
--------------------------------
<10> Burke R., Felfernig A., Goker M.H. Recommender Systems: An Overview // AI Magazine. 2011. Vol. 32. Iss. 3. P. 13 - 18.
1. The personalized nature of the recommendation system. In this case, the recommendations are intended to optimize the experience of a single user, and do not represent some kind of group solution.
2. The recommendation system is designed to help the user choose one of several options. As a rule, these options are already known in advance and are not generated individually.
It is interesting in the framework of this study to refer to generative information technologies being developed in Russia, for example, to YandexGPT 2, which was asked to generate its own definition of recommendation technologies.
As a result of the generation, YandexGPT 2 proposed the following option: recommendation technologies are methods and algorithms (processes, systems) that are used to analyze user preferences and behavior in order to provide personalized recommendations for various objects.
In general, the formulated approach of Russian legislation to the definition of recommendation technologies is quite consistent and interacts with other key concepts contained in the Law, in particular "information", "information technology".
Amendments to the Law on Information of the Russian Federation relate to prohibitions on the use of recommendation technologies that will violate the rights and legitimate interests of citizens and organizations, as well as violate the legislation of the Russian Federation in the process of providing such information.
In Articles 4 and 6 of the Regulations of the People's Republic of China on Recommendation Algorithms, in addition to the manifestation, as in Russian legislation, of the principle of legality, there is an expansion of requirements regarding the need to respect public morality and ethics, observe business and professional ethics, follow the principles of fairness and honesty, openness and transparency, scientific and reasonableness, reliability.
At the same time, certain paragraphs of the PRC's Regulations on Recommendation Technologies contain, in essence, requirements for internal security systems that must be fulfilled by providers of recommendation technology services.
For example, Article 7 lists a set of actions that should be aimed at ensuring the internal security of the functioning of algorithms, in particular verification, monitoring, security, improvement of management systems of recommendation technologies. Emergency response scenarios should also be provided, and Article 8 states the need for constant monitoring of the system's functioning mechanism.
In paragraph 2 of Article 10.2-2 of the Law on Information in the Russian Federation, a requirement is formulated on the need to inform users about the use of recommendation technologies, while the specific amount of information that will be disclosed, as well as the method of informing, is not defined in the article, they will be formulated by Roskomnadzor.
Special mention should be made of the draft Order of Roskomnadzor (at the time of writing) "On approval of the procedure for interaction of the Federal Service for Supervision of Communications, Information Technology and Mass Communications with the owners of information resources on which recommendation technologies are used" <11> (hereinafter - the Order of Roskomnadzor on interaction with owners of recommendation technologies). Paragraph 1.2 of the said Order gives Roskomnadzor the authority to send a request for access to software and hardware of recommendation technologies to assess the compliance of the use of recommendation technologies with the requirements of Articles 10.2-2 of the Law on Information in the Russian Federation.
--------------------------------
<11> Draft Order of Roskomnadzor "On approval of the procedure for interaction of the Federal Service for Supervision in the field of communications, Information Technology and Mass Communications with the owners of information resources on which recommendation technologies are used" // Federal Portal of draft regulatory legal Acts.
After the formulation of such powers, business entities had reasonable concerns due to the fact that the initial version of the draft implied the possibility of remote access to software and hardware from Roskomnadzor, which significantly increased the risks of leakage of information constituting intellectual property and protected by the trade secret regime.
At the moment, the possibility of remote access in the Roskomnadzor Order on interaction with owners of recommendation technologies in paragraph 3 has been replaced with the wording "... at the location of software and hardware or in another way agreed with the owner of the information resource."
These changes significantly reduce the risks of data leakage, but still maintain this threat, taking into account the human factor. For example, Karen Ghazaryan, Director General of the Internet Research Institute, holds a similar position, noting that there is still a risk of leaks from Roskomnadzor employees or its contractor.
--------------------------------
<12> IT companies argued with Roskomnadzor about providing access to their servers and code // Forbes.ru : official website.
For example, Article 29 of the Regulations on Recommendation Algorithms of China separately establishes the requirement for the institution and personnel who participate in the security assessment, supervision and inspection of algorithm recommendation services to maintain the confidentiality of personal information and trade secrets.
In this regard, a valuable conclusion deserves attention, reflected in the article by Y.S. Kharitonova and T. Yang "Recommendation systems: development of legal approaches in Russia and China", in which the authors note that in striving for transparency and control over recommendation algorithms, it is necessary to strive to achieve a balance between innovative algorithmic solutions and explicit and implicit violations principles of fair application of recommendations <13>.
--------------------------------
Kharitonova Yu.S., Yan T. Recommendation systems: development of legal approaches in Russia and China // Entrepreneurial law. 2022. N 2. pp. 50-58.
Also noteworthy from the point of view of the practical implementation of the legal mechanism being built is the position of the Chairman of the Council of the Digital Economy Development Fund. German Klimenko <14> notes that checking the operation of the recommendation algorithm requires researching a large amount of product code, which will require a lot of time and highly qualified programmers. In this regard, indeed, in practice, Roskomnadzor may face big problems when performing product code checks related to the operation of recommendation algorithms.
--------------------------------
<14> Kharitonova Yu.S., Jan T. Decree. op.
For comparison, Article 23 of the PRC Regulations on Recommendation Algorithms should be mentioned, which enshrines a different, more systematic approach. It is planned to create a classification of algorithms and a security management system, while evaluating the algorithm based on the ability to influence public opinion, content categories, user scale and data that are processed by recommendation technologies. A hierarchy will be built based on this data.
We also see demands for transparency. For example, the VKontakte social network has formed Rules for the use of recommendation technologies of the VKontakte social network and, in compliance with the requirements that entered into force on October 1, 2023, it also sent notifications within the social network indicating a link to familiarize itself with these Rules. According to these Rules, recommendations are built on the basis of data received from the marks "I like", "This is not interesting", "Share", the frequency of viewing content, etc.
Thus, the analysis of the provisions of the legal regulation of recommendation algorithms in Russia and China demonstrates the active development of the legal framework in both countries, while in China the legal formalization of the issue under consideration began earlier and is currently more systematic and detailed.
The system of requirements imposed, including on business entities, in practice requires the creation of an internal compliance system that will monitor and monitor the operation of information technologies in order to prevent violations of the law.
By recommendation technologies, the author suggests understanding information technologies, the functioning of which is based on personalized user preferences.