The problem of legal assessment of the essence of scientific and educational texts from the perspective of the role and place of the author in the generative content of neural networks

Belikova Ksenia Michailovna ORCID: 0000-0001-8068-1616 Doctor of Law Professor of the Department of Entrepreneurial and Corporate Law, Kutafin Moscow State Law University 125993, Russia, Moscow, Sadovaya-Kudrinskaya str., 9 BelikovaKsenia@yandex.ru Other publications by this author

The constant development of digital technologies and artificial intelligence and their introduction into education is mandatory for most leading universities in Russia and abroad, since technologies based on artificial intelligence (for example, neural networks, machine learning, etc.) give a new impetus to the development of universities and educational institutions. However, when implementing AI, it is important to ensure the solution of a number of tasks, such as supporting, rather than replacing, AI teachers, teachers who use AI in their activities; ensuring the confidentiality and security of data of teachers and students; solving technical, legal, ethical and other problems that have arisen in connection with the development of generative AI, including clarifying the role, place and the very concept of the author and authorship, including from the point of view of law.

In this format, the purpose of the study is to scientifically comprehend and analyze legal approaches to regulating the work of neural networks and generative artificial intelligence in the context of scientific texts created by them/with its help from the perspective of the possibilities of solving the problem of trust in such results and the use of AI in education, while simultaneously assessing the pros and cons of AI support provided to people in general and participants in educational in particular, taking into account the provisions of current copyright law and approaches to recognizing the authorship of AI in foreign legal systems (prepared based on the report and presentation at the World Congress "Theory of Systems, Algebraic Biology, Artificial Intelligence: Mathematical foundations and Applications", June 26-30, 2023, Moscow, etc. (https://congrsysalgbai.ru/ru /) (URL: https://youtu.be/2cs1uAhWaPI (date of request: 08.10.2023)). Some answers will be given to the questions of how to use artificial intelligence to improve students' knowledge, how, based on this, the government should strengthen oversight of the risks of artificial intelligence so that the use of AI in education is "effective" and "safe", etc.

Currently, the issue of neural networks generating various content, including scientific texts, is becoming the subject of attention at scientific conferences (e.g., a report by PhD, Assoc. Pavlova A.A. Kazan (Volga Region) Federal University "Scientific texts created using artificial intelligence: current state and prospects of development" from the X Perm readings on the methodology of civil studies (May 26-27, 2023, PSNIU, Perm), in publications (e.g., ^{[23, C. 3-6]}, Yankovsky R.M. Is artificial intelligence capable of writing an article in a law journal? 27.03.2023. URL: https://zakon.ru/publication/igzakon/10319?ysclid=libg15gofu 26021959 (date of access: 30.05.2023)), in the business community (for example, the webinar of the Anti-Plagiarism system on 05/16/2023 "ChatGPT: Pandora's box or porridge from an axe" (author - Chekhov Yu.V.). URL: https://antiplagiat.ru/training/16-05-2023 /?ysclid=libgfu9vnk488745193 (accessed: 30.05.2023).

In this regard, we have chosen this issue for coverage because the possibilities offered by neural networks (e.g. ChatGPT, YaGPT, etc.) give rise to a number of technical, legal, ethical and other problems. So:

1) according to the information from the named webinar of the Anti-Plagiarism system, ChatGPT writes diplomas: according to a student of RSUH, he defended a diploma written using ChatGPT. The anti-Plagiarism system, after the introduction of a new module on May 02, 2023, may call a document suspicious and assume its text is generated, but today, when working with such documents, it cannot be proved that the text is artificial, and this is a technical problem;

2) another problem is that the regulatory framework for regulating the functioning of neural networks is so insignificant today that it actually allows you to use this technology in any field without restrictions, which can form a moral and ethical problem;

3) the next problem is the generation of fictional reality by neural networks based on the generation of false information (content), operating with non-existent facts, normative legal acts, events, which creates legal uncertainty, people and society become disoriented.

What features of neural networks contribute to the emergence of such problems? Researchers note [1. pp. 58-63], firstly, the "ability" of neural networks to change their state in response to the effects of the external environment, to respond to the effects of external stimuli is not quite straightforward, that is, to isolate through various interference and process only the necessary information. Such a skill is of key importance for a neural network system while working in real conditions and distinguishes this system from a computer in which the principle of generality is applied, that is, the system is not able to adapt to external influences, but only performs those operations that are embedded in it by the programmer. Secondly, the "ability" to reproduce on the basis of distorted input data at the output the ideal essence of what was provided at the input in a distorted form, in our opinion, is a curious consequence of the first feature, interesting from the point of view of philosophy.

Regarding each of these problems and the ways to solve them, the following can be said.

There are a number of Russian initiatives on the issue of legal regulation of neural networks, for example, that:

1) "... a symbiosis of self-regulation and regulation of public relations in the field of neural networks in various spheres of public life can become a model of legal regulation of neural networks. Self-regulation ... should consist in building a system of technical regulations and standards (GOST and ISO system), and at the regulatory legal level it is necessary to consolidate the concept of neural networks, the legal status of their subjects, principles, limitations and limits of regulation, areas in which the use of neural networks is prohibited, limits in training.", and further it is required "...to include in the federal draft "Regulatory regulation of the digital environment" of the national program "Digital Economy" (Approved by the Presidium of the Presidential Council for Strategic Development and National Projects (Protocol No. 16 dated 12/24/2018) the task of legal regulation of these relations." [2. pp. 235-237],

2) special ethical and legal principles and requirements should be developed that apply ... to the development of algorithms, devices and solutions based on artificial intelligence and to the machine learning process, including self-learning, ... taking into account the specifics of this process [3. pp. 219-231];

3) it is necessary to create a list of companies that have the right to train neural networks, or create a network collaboration of leading companies in the industry that will work within the framework of the Chesborough open innovation strategy [4. pp. 51-63; 5. pp. 58-83; 6. pp. 174-181; 7. pp. 182-190] in order to strengthen the potential of development companies and to avoid the danger of concentration in their hands of copyrights to objects created by neural networks and/or using neural networks by such companies, and, as a result, monopolization of the industry and further prohibition of other industry participants – people - creators of any works (text, images, music, etc.) to use their works as repetitive (plagiarizing) works created earlier by a neural network, allowing human creators of works to use artificial intelligence to create the distinctive ability of works. One of the possibilities for creating the latter, a field of legal certainty and security in the conditions provided by networking, is a blockchain based on the use of distributed registry technology, in which any interested person will have access to information about the history of entering an intellectual property object into the registry and about its exclusive rightholder. At the same time, information will be transmitted with identical content, since the authorship of any changes in the registry is fixed during recording and subsequently (when distributing copies) is easily established and verified [8. pp. 381-382].

What are these initiatives based on?

Among the acts creating vectors of the regulatory impact of neural networks and artificial intelligence in Russia, one can specify, for example, the following:

- Decree of the President of the Russian Federation dated 10.10.2019 No. 490 "On the development of artificial intelligence in the Russian Federation" (approved by the "National Strategy for the development of artificial Intelligence for the period up to 2030") (URL: https://www.garant.ru/products/ipo/prime/doc/72738946 / (date of request: 14.10.2022));

- The Federal Artificial Intelligence Project (hereinafter referred to as the AI FP) is valid until the end of 2024 and provides for a set of multidisciplinary tasks fixed by the National AI Development Strategy (Approved by Appendix No. 3 to the protocol of the Presidium of the Government Commission on Digital Development, the Use of Information Technologies to improve the quality of Life and business Conditions dated 08/27/2020 17. URL: https://www.consultant.ru/document/cons_doc_LAW_398627 / (date of access: 05/11/2023));

- Forecast of socio-economic development of Russia until 2030 (URL: http://static.government.ru/media/files/41d457592e04b76338b7.pdf (accessed: 05/28/2023)), which involves the development of semantic networks – distributed networks with independent nodes and adaptive routing between them in terms of working with content (Semantic Web);

- earlier, the Order of the Federal Customs Service of Russia dated March 27, 2012 No. 575 "On export control of dual-use goods and technologies that can be used in the creation of weapons and military equipment and for which export control is carried out" (URL: https://base .garant.ru/58055391/?ysclid=lidb29hku2320945011 (date of reference: 30.05.2023) (expired in 2017) defined such a concept as a "neural computer", which meant "a computing device designed or modified to simulate the behavior of a neuron or a set of neurons, for example, a computing device characterized by the ability of equipment to modulate the weight and number of mutual connections of a set of computing devices components based on previous information";

- GOST standards, for example, GOST R 70462.1-2022/ISO/IEC TR 24029-1-2021 is the national standard of the Russian Federation. Information technology. Artificial intelligence. Evaluation of the robustness of neural networks. Part 1. Overview (date of introduction 2023-01-01) (URL: https://docs.cntd.ru/document/1200193906 ?ysclid=lktl0f829l320002588 (accessed: 07/26/2023)) and others.

However, Article 1257 of the Civil Code of the Russian Federation (Civil Code of the Russian Federation; Part Four of the Civil Code of the Russian Federation dated December 18, 2006 No. 230-FZ) still acts as the regulatory norm. // Federal Law of the Russian Federation dated December 25, 2006 No. 52 (part I) of Article 5496), according to which "the author of a work of science, literature or art is recognized as a citizen whose creative work it was created" [9. pp. 7-24; 10. pp. 47-61].

The laws of such countries of the British Commonwealth as the United Kingdom, Australia and Canada, for example, are still following this path:

- in the UK, the author of a literary, dramatic, musical or artistic work created using a computer (literary, dramatic, musical or artistic work which is computer-generated) is still understood by the Copyright, Industrial Designs and Patents Act 1988 in art. 9(3) as a person who has taken the necessary measures to create it;

- in Australia, according to the Copyright Act of 1968, the author can only be a "qualified person" (a citizen or resident of Australia – Article 32), which excludes the possibility of a neural network to be the author of a work;

- In Canada, according to the Copyright Act of 1985, the author of a work can only be a citizen, a subject, or a person usually residing in the country of the Contract (art. 5(1)(a) and further). According to the references to the opinion of the Government of Canada in 2018, the country of the Treaty is defined as a country of the Berne Convention, a country of the World Copyright Convention or a member of the World Trade Organization (Government of Canada, 2018. // Deb Weston. The Basics of Understanding Copyright in Canada. January 13, 2019. URL: https://heartandart.ca/the-basics-of-understanding-copyright-in-canada / (date of access: 01/28/2024)).

An example of such an understanding being implemented by the courts is the decision of the Australian court in the case IceTV Pty Limited v Nine Network Australia Pty Limited HCA 14 dated April 22, 2009 on the refusal to protect the copyrights of artificial intelligence (Agenda Media Group: world practice of legal regulation of artificial intelligence. 06.06.2023. URL: https://www.sostav.ru/publication/agenda-media-group-ugolovka-za-dipfejki-60946.html?ysclid=lkjublamqc656426741 (date of access: 01/28/2024); Vasilyeva A. On the issue of copyright in artificial intelligence. February 1, 2023. URL: https://www.garant.ru/article/1605912 / (date of access: 01/28/2024)).

At the same time, there are works (e.g., Morkhat P.M. The legal personality of artificial intelligence in the field of intellectual property law: civil law problems. Dis... Dr. jurid. Sciences. - M., RGAIS, 2018. pp. 181-221), which analyze and qualify the issues of action and qualification of artificial intelligence units in the context of law from the position of intellectual property as an author, co-author, employee or tool.

In line with some of the approaches outlined in such works, there are precedents in a number of countries when, by court decision, artificial intelligence is endowed with a certain amount of copyright. For example, this is the decision of the People's Court of Nanshan District (Shenzhen, Guangdong Province) in 2019 in the case of Shenzhen Tencent Computer System Co., Ltd. v. Shanghai Yingxun Technology Co., Ltd. ( ? ? ? ? ? ? ? ? . URL: https://ru.chinajusticeobserver.com/law/x/2019-yue-0305-min-chu-14010 (date accessed: 28.01.2024), under which the copyright to works created by artificial intelligence (in this case software plaintiff Dreamwriter) may be protected in accordance with the laws of China and the works of the people, although the copyright Law of China 1990 (as amended by 2020) as before "ties" the right of authorship of the work to such subjects as Chinese citizens, legal persons or unincorporated organization; and, in the case of foreigners or stateless persons – 1) in accordance with the agreement concluded between China and a country to which the authors belong, or in which they have their habitual residence or 2) in accordance with an international Treaty between the two countries, etc.

However, China recently went on: April 11, 2023, the Office for cyberspace China (Cyberspace Administration of China) to promote the healthy development and standardized technology generative artificial intelligence based on the Laws of the PRC, about cybersecurity, about data security and the protection of personal information [15. P. 267-268] and administrative regulations posted on its website http://www.cac.gov.cn/ developed by the Project "measures for the management of services generative artificial intelligence" (2023?04?11? URL: http://www.cac.gov.cn/2023-04/11/c_1682854275475410.htm (date accessed: 28.01.2024)) and opened it for public comment. The Draft states that the Measures relate to products related to technologies that generate text, images, sound, video, codes and other content based on algorithms, models and rules (Article 2 of the Draft). Among the measures proposed for implementation to prevent disorientation of people consuming content generated by artificial intelligence (neural network), mandatory labeling of generated images, videos and other types of content is proposed.    Thus, according to Article 8 of the Draft, when labeling manually generated AI content, suppliers of such content should develop clear, specific and workable labeling rules that meet the requirements of these Measures, conduct the necessary labeling training and sampling to verify the correctness of the labeling.

According to Articles 10 and 18 of the Draft, providers must explain the cases and ways of using the products and services they provide that contain content generated by the neural network, as well as take appropriate measures to prevent users from relying too much on or condoning content creation.

In addition, providers should direct and draw the attention of users to the need for scientific understanding and rational use of AI-generated content, explain to them not to use the generated content to harm the image, reputation and other legitimate rights and interests of others, as well as not to participate in commercial advertising of any goods (services) or their inappropriate marketing (see, for example, Ermakova I.V. Legal regulation of relations in the field of advertising, containing a comparison of the advertised product with competitors' products, in the Russian Federation and Germany. Dis...cand. Jurid. Sciences. – M.: RUDN, 2017).

It is worth remembering here that this warning is being made in the context of the currently rapidly developing social commerce (i.e., commerce developed on social networks), which is understood as electronic sales (e-commerce) using social networks to support social interactions and user participation, to mediate online purchase and sale of goods and services where social networks are considered as platforms where people can connect online, receive advice from trusted users (including bloggers and influencers), find goods and services, and then buy them, thus making such channels channels functioning on the basis of a word-of-mouth strategy [8. P. 235].

At the same time, the Draft Measures in Articles 4 and 12 fix the requirements for the content generated by AI, which can be contained in products and services – the results of such generation:

1) it must embody the basic socialist values and must not contain anything aimed at: undermining state power, overthrowing the socialist system, incitement to separatism; undermining national unity; propaganda of terrorism and extremism, propaganda of national hatred, violence, obscene pornographic information, false information and content that may disrupt economic and social The order,

2) It must be truthful and accurate, and measures must be taken to prevent the creation of false information,

At the same time, measures should be taken to prevent discrimination based on race, ethnicity, beliefs, nationality, region, gender, age, occupation, etc. in the process of developing algorithms, selecting training data, generating and optimizing models, as well as providing services. It is worth saying here that this kind of concern is shared in other countries (South Africa, Brazil from among the BRICS countries, for example [9. pp. 1-21; 10. pp. 1-23]).

In addition, such content should be built with respect in mind.:

1) intellectual property rights and business ethics, therefore, as a result of content generation, the advantages provided by algorithms, data accumulation and the use of platforms should not be used to carry out unfair competition.

It is interesting to recall, on the one hand, a number of Russian novels of this kind, for example, the Federal Antimonopoly Service of Russia (hereinafter referred to as the FAS RF) has been developing a "fifth" antimonopoly package for some time, entirely aimed at regulating the digital environment. On July 10, 2023, the President of the Russian Federation signed Federal Law No. 301-FZ dated 07/10/2023 "On Amendments to the Federal Law "On Protection of Competition" (Federal Law of the Russian Federation, July 17, 2023, No. 29, Article 5319) (hereinafter referred to as the Law), which is part of the "fifth antimonopoly package" (the second bill (See: Sokolovskaya E. The State Duma is preparing for adoption a bill to strengthen responsibility for the use of digital algorithms in cartels. 12/29/2022. URL: https://www.pgplaw.ru/analytics-and-brochures/alerts/gosduma-gotovit-k-prinyatiyu-zakonoproekt-ob-usilenii-otvetstvennosti-za-ispolzovanie-tsifrovykh-alg / (date of appeal: 07.12.2023)) from the fifth antimonopoly package No. 160278-8 "On Amendments to the Code of Administrative Offences of the Russian Federation" – clarifies administrative liability for violations related to manifestations of monopolistic activity in digital commodity markets, and was adopted by the State Duma in the first reading). First, it defines the concept of "network effect"  as a property of the commodity market (commodity markets), in which the consumer value of a program (set of programs) for electronic computers in information and telecommunications networks, including on the Internet, providing transactions between sellers and buyers of certain goods (digital platform), varies depending on changes in the number of such sellers and customers. Thus, the new regulation will affect commodity markets where transactions are made using digital platforms, in connection with which, the antimonopoly authority is required to establish the presence of network effects and assess the ability of an economic entity owning a digital platform and ensuring through its use transactions between other persons acting as sellers and buyers of certain goods, to provide a decisive influence on the general conditions of circulation of goods in such a commodity market, in which transactions are carried out through the use of a digital platform. Secondly, the Law on Protection of Competition has been supplemented by Article 10.1, which provides for the prohibition of monopolistic activities by persons owning a digital platform and ensuring through its use transactions between other persons acting as sellers and buyers of certain goods, subject to a combination of the following conditions: the network effect gives an economic entity the opportunity to have a decisive influence on the general conditions of circulation of goods on the market a commodity market in which transactions are carried out through a digital platform (to eliminate other household entities from this commodity market, to make it difficult for them to access this commodity market); the share of transactions made between sellers and buyers through a digital platform exceeds 35% of the total volume in value terms of transactions made in the corresponding commodity market; revenue the economic entity exceeds more than 2 billion rubles in the last calendar year. At the same time, the final version of the Law provides for the right of an economic entity to provide evidence that its actions (inaction) can be recognized as permissible in accordance with the requirements of Part 1 of Article 13 of the Law on Protection of Competition (paragraph 2 of Article 10.1).

On the other hand, foreign experience in combating the accumulation of user data by digital platforms and their misuse is interesting and useful – for example, the consideration by the European Commission and the Federal Anti-Cartel Office of Germany (Federal Cartel Office, FCO) of Amazon's actions to determine whether it uses seller data to gain advantages when selling its products directly on the German market [pp. 140-141, 156-157];

2) the legitimate interests of other persons, preventing harm to the physical and mental health of others, damage to image (portrait) rights, reputational rights and privacy, as well as preventing violations of intellectual property rights. It is prohibited to illegally acquire, disclose and use personal information, violate privacy and trade secrets.

Here, distracting from the issues of privacy and prevention of violation of intellectual property rights, which are also, of course, important, we believe it is necessary to focus on the fact that an important aspect of future work on the prevention of harm to health is, of course, the promotion of physical education in order to minimize risks to health and well-being, as a physical - deteriorating eyesight, weak backs, the development of varicose veins from unbalanced sitting and lack of movement, etc., and psycho-psychological - "sticking" in a computer (tablet, smartphone), separation from real life, as a result of which a false idea of reality is formed, etc.

Although there are proposals for the protection of intellectual property rights, but in another legal order – the Japanese one (see Agenda Media Group: world practice of legal regulation of artificial intelligence. 06.06.2023. URL: https://www.sostav.ru/publication/agenda-media-group-ugolovka-za-dipfejki-60946.html?ysclid=lkjublamqc656426741 (accessed: 07/26/2023)). The country proposes to limit the protection of works created by artificial intelligence and treat them as a trademark, bringing the relevant legal regulation into the field of unfair competition.

With this approach, the rights to digital works will belong to the creator of the corresponding algorithm, which, according to the authors of such an idea, will help limit the protection of popular works – the results of generation and remove them from the scope of copyright.

It is worth saying that the idea of labeling AI-generated content is supported by many law enforcement agencies and supranational associations.

So, from a blog post by OpenAI (Moving AI governance forward. OpenAI and other leading labs reinforce AI safety, security and trustworthiness through voluntary commitments. July 21, 2023. URL: https://openai.com/blog/moving-ai-governance-forward (accessed: 01/28/2024); Annie Bronson. OpenAI, Google and five other major IT companies have agreed to introduce watermarks for AI-generated content. July 23, 2023. URL: https://habr.com/ru/news/749804 / (accessed: 01/28/2024)) it follows that seven large IT companies (OpenAI, Microsoft, Google, Meta, Amazon, Anthropic and Inflation) have committed themselves to "develop watermarks for audio and visual content", with the exception of virtual assistant voices, for example, as well as "tools or APIs to determine whether a particular piece of content was created using AI." It is unclear exactly how watermarks will work, but most likely they will be embedded in the content so that users can trace its origin.

Developers and regulators expect this measure to help curb the spread of fake news and misinformation. The fact is that a number of cases that caused a lot of noise objectively pushed companies and the Administration to such a decision. For example, the Midjourney image generator was used to create fake images of Donald Trump's arrest: journalist Eliot Higgins, according to her, helped create fake images using artificial intelligence and posted fake photos on her Twitter account. If the watermark had been available then, she would not have faced the consequences of what, according to her, was not an attempt to be clever or deceive others, but just to have fun with Midjourney (see Makarychev M. Fake pictures of Trump's "forcible arrest" by the police appeared on the network. 03/21/2023. URL: https://rg.ru/2023/03/21/v-seti-poiavilis-fejkovye-snimki-silovogo-aresta-trampa-policejskimi.html (date of access: 01/28/2024); Yuri. OpenAI, Google will put watermarks on AI-generated content to prevent deepfakes. July 25, 2023. URL: https://vc.ru/s/1420039-neyroseti/770804-openai-google-postavit-vodyanye-znaki-na-kontent-sozdannyy-ii-chtoby-predotvratit-dipfeyki (date of application: 01/28/2024)). In this regard, the White House stated that the watermark would allow "creativity with AI to flourish, but reduce the risk of fraud and deception" (Ibid.).

Google also said that in addition to watermarking, it also integrates metadata and "other innovative methods" to verify the authenticity of information.

These companies also made a number of other commitments: they promised to conduct both internal and external testing of AI models before their release; they stated that they would invest more in cybersecurity and share information with the industry in order to reduce the risks from AI, starting with bias and ending with the use of AI in the development of weapons.

This idea is being developed by these companies in line with the Biden-Harris Administration initiative, on the basis of which the Administration seeks voluntary commitments from leading artificial intelligence companies to manage AI-related risks (Fact Sheet: Biden-Harris Administration Secures Voluntary Commitments from Leading Artificial Intelligence Companies to Manage the Risks Posed by AI. July 21, 2023. URL: https://www.whitehouse.gov/briefing-room/statements-releases/2023/07/21/fact-sheet-biden-harris-administration-secures-voluntary-commitments-from-leading-artificial-intelligence-companies-to-manage-the-risks-posed-by-ai / (date of access: 01/28/2024)). This initiative is revealed and implemented through the ideas of safety & security and trust, as well as the Decree of the President of the United States dated February 11, 2019 on preserving American leadership in the field of artificial intelligence ^[16], prescribing the creation of the "American Artificial Intelligence Initiative" within the framework of the current regulatory window, designed to stimulate development and regulation of AI and prescribing to the federal agencies implementing this Decree, which: conduct fundamental research and development in the field of AI; develop and implement applications of AI technologies; provide educational grants; regulate and make recommendations on the use of AI technologies. At the same time, the Co-Chairs of the Special Committee of the National Council for Science and Technology (NSTC Select Committee) determined that this Initiative should be implemented through a coordinated strategy of the federal government based on five principles calling for:

(1) to promote technological breakthroughs in the field of artificial intelligence within the federal government, industry and academia;

(2) encourage the United States to develop appropriate technical standards and reduce barriers to safe testing and implementation of AI technologies in order to ensure the creation of new AI-related industries and its implementation in modern industries;

(3) to train current and future generations of American workers in the development and application of artificial intelligence technologies to prepare them for today's economy and the jobs of the future;

(4) strengthen public trust in AI technologies and protect civil liberties, privacy, and American values in their application in order to fully realize the potential of such technologies for the American people;

(5) contribute to the creation of an international environment that supports American AI research and innovation and opens markets for American AI industries, while protecting the U.S. technological advantage in AI and protecting critical U.S. AI technologies from acquisition by their strategic competitors.

So, from a safety perspective, companies are required to ensure the safety of their products before presenting them to the general public, which means that: 1) testing, including external, the security of artificial intelligence systems and their capabilities; 2) assessing their potential biological and social risks, cybersecurity risks, and 3) publishing the results of these assessments.

From a security perspective, companies are required to create systems that put security first. This means protecting their models from cyber and insider threats and sharing best practices and standards to prevent abuse, reduce risks to society and protect national security.

From a position of trust, companies are obliged to do the right thing towards society and deserve the trust of people. This means that it will be easy for users to determine whether audio and visual content is in its original form or has been modified or generated by artificial intelligence. This also means ensuring that technology does not contribute to bias and discrimination, strengthening privacy protection and protecting children from harm. Finally, it means using artificial intelligence to solve the most serious problems of society (from cancer to climate change) and managing the risks associated with its work in such a way that the benefits provided by its use are fully realized (see Ensuring Safe, Secure, and Trustworthy AI. URL: https://www.whitehouse .gov/wp-content/uploads/2023/07/ Ensuring-Safe-Secure-and-Trustworthy-AI.pdf (accessed: 01/28/2024); Moving AI governance forward. OpenAI and other leading labs reinforce AI safety, security and trustworthiness through voluntary commitments. July 21, 2023. URL: https://openai.com/blog/moving-ai-governance-forward (date of application: 01/28/2024)).

Interesting, but also dangerous, in this regard is the fact that now, according to Google's developed policy, reflected in the Webmaster's Guide (see Google Search Essentials (formerly Webmaster Guidelines). URL: https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&cad=rja&uact=8&ved=2ahUKEwiLg_2f29aAAxXZJxAIHV4qAlQQFnoECBsQAQ&url=https%3A%2F%2Fdevelopers.google.com%2Fsearch%2Fdocs%2Fessentials&usg=AOvVaw1nPRmqmfT1NANvtzOMJ7fZ&opi=89978449 (accessed: 01/28/2024)), it equates AI content generation to spam.

What creates concerns? According to the "Google Web Search Spam Rules" section of this Guide, Google's spam rules help protect users and improve the quality of search results.

How? In order to be allowed to be displayed in search results (whether it is web pages, images, videos, news content or other materials that Google finds on the Internet), the content must not violate the general Google Search rules or the spam rules presented in this section. These rules apply to all Internet search results, including on Google resources.

What is Google actually doing? In order to identify content and behavior that violate the rules, automatic systems have been developed, if necessary, verification is carried out by Google specialists who can take measures manually. Sites that violate Google's rules may drop lower in the list of search results or not be shown there at all. Imagine your website that you want to promote, containing partially generated content. Will you achieve progress with this approach, which is followed by Goggle?

What do they say about this in the company's administration? This topic was raised as part of another conversation between John Mueller, a leading Google analyst who deals with search quality, and webmasters on the Google Search Central YouTube channel. The question concerned the content created by GPT-3. No matter what tools are used to create it (if you're just rearranging words or looking for synonyms, or doing translation tricks that people have used before… we will consider this spam), content written by machines is considered to be automatically created (see Google equates to spam the generation of content using AI. URL: https://seo-aspirant.ru/ai-generaciya-kontenta (date of application: 08/12/2023)).

Examples of automatically created content from Google include, among other things, text created automatically as a result of the selection of synonyms; content copied from various web pages without adding unique information, etc.

In line with this understanding, the danger becomes even wider, because with this understanding, the paraphrasing forced from a conscientious author by the Russian Anti-Plagiarism system according to the current rules of its algorithm [17. pp. 62-104; 18. Pp. 1-11] becomes automatically created content, which in reality is not the case at all.

When asked if Google can understand the difference between content written by humans and content written by robots, D. Mueller replied, "I can't say that. But if we see that something is generated automatically, then the anti-spam team can definitely take action on this." and "From our point of view, we still consider this to be automatically generated content. I think that over time, maybe it will evolve and become more of a tool for people. It's like how you use machine translation as the basis for creating a translated version of the site, but you're still working on it manually. Perhaps over time, these AI tools will evolve in such a way that you will use them to write texts more efficiently or to make sure that you write correctly, such as spelling and grammar checking tools, which are also based on machine learning. But I do not know what the future holds for us." (see: Google equates the generation of content using AI to spam. URL: https://seo-aspirant.ru/ai-generaciya-kontenta (date of application: 01/28/2024))

At the same time, another interesting fact about the work of Google's dig is that according to the new version of Google's privacy policy, all content published on the Internet can be used by it to train its Bard neural network (see: Kildyushkin R. Google has decided to assign all content on the Internet. July 4, 2023. URL: https://turbo .gazeta.ru/tech/news/2023/07/04/20804846.shtml (date of application: 01/28/2024)), which, in our opinion, may pose a threat of various violations (for example, image rights, etc.).

In the European Union (EU), it was proposed to oblige services to label content created by neural networks. One of the European examples that could raise concerns about the generation of AI content is as follows. Researcher Bob Sturm and colleagues from the Royal Institute of Technology in Stockholm came up with the band O Conaill Family and Friends, which allegedly performs music in the style of Irish-Celtic folk, and posted their first album on Soundcloud called Let's Have Another Gan Ainm, writes inc.com . After that, the experts gave the recordings to critics and others to listen to. All of them highly appreciated the work of the folk group and predicted a great future for it. And later they refused to believe that they had been tricked, and a wonderful album was created by a computer (see: Wiltovsky M. Professional musicians could not distinguish the creation of a computer from the music of a living composer. October 10, 2018. URL: https://www.sb.by/articles/melomany-ne-smogli-otlichit-tvorenie-kompyutera-ot-muzyki-zhivogo-kompozitora.html (date of application: 01/28/2024)). In this context, Vice-President of the European Commission Vera Jourova and Commissioner for Internal Market Thierry Breton met with representatives of more than 40 organizations, including TikTok, Microsoft, Meta, etc., who signed (among the signatories there is no Twitter) the voluntary Code of Practice on combating disinformation of the European Union 2022 ^[19] Labeling should be applied to all materials created with the help of artificial intelligence, be it audio, video, text or images. This voluntary tool can be integrated into one of the acts that the relevant EU bodies and institutions are working on - on digital services or on artificial intelligence. In addition, companies that integrate generative AI into their services - for example, Microsoft's Bing chat and Google's Bard - must provide security measures so that their products are not used to create fakes. "Today, such content is mostly a game, but tomorrow you can see fake porn with a presidential candidate," a source told Politico Playbook (see: Khabidulina E. In the European Union, it was proposed to oblige services to label content created by neural networks. // Forbes. June 5, 2023. URL: https://forbes-ru.turbopages.org/forbes.ru/s/tekhnologii/490383-v-evrosouze-predlozili-obazat-servisy-markirovat-sozdannyj-nejrosetami-kontent (date of application: 08/12/2023)).

These ideas and practical steps for self-regulation are in line with the approach practiced in another field - the protection of human rights and business entities from unlawful (without)the actions of the public administration in Japan, where the protection of rights violated by the public administration is the prerogative of public organizations and the issue of social responsibility of business, that is, the principle "behave correctly - everything will be fine" ^[20] (see: The Business Commissioner of Transbaikalia discussed the prospects for interaction with the Consul General of Japan. 03/21/2019. URL: https://ombudsmanbiz.75.ru/novosti/190941 (date of access: 30.04.2023)).

Similarly, in Russia, legislators are making proposals to label content created using neural networks. Experts from the Russian Technological University MIREA suggested doing this with the help of graphic signs (see: The State Duma proposed labeling content created using neural networks. May 29, 2023. URL: https://turbo.ria.ru/20230529/kontent-1874878008.html (date of application: 08/12/2023)).

At the same time, there is another idea – to license music created with the help of AI. Google and Universal Music are working on such a solution due to the increased number of musical fakes – modern musical works created by imitating the voices of famous performers, including the deceased, who, like probably their descendants, did not consent to such use of their voice. For example, Frank Sinatra's voice was used in the hip-hop version of the song "Gangsta's Paradise", and Johnny Cash's voice was used in the popular single "Barbie Girl". Warner Music also intends to participate in the project. It is assumed that thanks to licensing, a tool will appear that will give anyone the opportunity to legally create tracks using AI, while singers and performers themselves will decide whether to consent to such use of their voice or not, and if they agree to receive royalties (Sostav.ru . Music created with the help of AI will be licensed. August 9, 2023. URL: https://www.sostav.ru/publication/muzyku-sozdannuyu-s-pomoshchyu-ii-nachnut-litsenzirovat-62330.html (date of application: 01/28/2024)).

As for the question of whether AI is the author of a work, or is it just a mechanism for writing a work, similar to a ballpoint pen, for example, then:

- in a number of reports during the World Congress "Theory of Systems, Algebraic Biology, Artificial Intelligence: mathematical foundations and applications" (June 26-30, 2023, Moscow, Russia; URL: https://congrsysalgbai.ru/ru / (date of reference: 08/12/2023)) (for example, D.V. Ushakova (Institute of Psychology of the Russian Academy of Sciences) "Architecture of the human cognitive system and artificial intelligence", K. Anokhina (Institute for Advanced Brain Research, Lomonosov Moscow State University) "Consciousness in neural networks", etc.) indicated, in particular: that "consciousness can be considered as a specific process in specifically organized structures (e.g., a neural hypernet)" (K. Anokhin), in the context of the architecture of cognition and the action of two systems of cognition and information processing in humans – conscious, voluntary and subconscious (insight, implicit decisions), it was noted that "The general structure of cognitive activity can be carried out by different elements – based on carbon, silicon, etc. and they can interact" (D.V. Ushakov) and etc.,

- in an interview with Alexander Kuleshov, Rector of the Skolkovo Institute of Science and Technology, Chairman of the Scientific Council of the IPPI RAS, Academician of the Russian Academy of Sciences, as part of the joint media project of Ogonek magazine with the A.A. Harkevich Institute of Information Transmission Problems of the Russian Academy of Sciences "Mathematical Walks" ("We already live in a new reality" Mathematician Alexander Kuleshov — about an artificial neural network, which already knows how to think about the information field. August 25, 2021 https://cont.ws/@sage/2069866 (date of application: 01/28/2024)) The academician says: "Until now, there is no mathematical explanation for the work of neural networks... we do not understand the laws of physics that are involved in their work.";

- at the same time, people, preparing scientific works for publication, of course, are engaged in compilation, since it is impossible to create a new one without relying on the old, even when it comes to inventions [17. pp. 62-104]. Therefore, in D. Nimmer's work "Copyright in the Scrolls of the Dead Sea. Authorship and originality" 2001 [21. P. 14], which investigated in detail the issues raised in the title, speaking about the amount of originality (Quantum of Originality) of the author's work, referring to the 1991 case Feist Publics Inc. v. Rural Tel. Serv. Co. (499 U.S. 340, 362 (1991). Cit. according to: 21), for example, notes that "the threshold for copyright protection is low. (...the threshold for copyright protection is low)" and wonders: "What could be a greater proof of originality than exclusivity (distinctiveness)...a contribution? (What greater proof of originality could there be than the distinctiveness of … contribution?),

- whereas in legislation, the text that is the result of the work of a neural network is designated by the term computer-generated / computer-generated, since, apparently, the legislator himself does not fully understand how best to mark the works obtained as a result of the work of neural networks today.

In this context, it seems that the help of artificial intelligence for humans may be very doubtful for the following reasons:

1. Today, a person is not able to separate the generated information from the real one on his own, even in the sphere previously attributed to a purely human prerogative, namely, creativity. This applies both to the creation of musical and artistic works, and to the creation of new scientific information based on an acquired, meaningful and developed human knowledge complex.

2. Human competition with AI is already practically impossible, since AI is the carrier and active user of a huge array of information obtained from numerous sources, which one person, even with encyclopedic knowledge, does not own and is not able to master in the foreseeable period of time, if only because he has a significant number of physiological limitations and needs.

3. Neural networks can cause significant harm to a person personally and to the human population as a whole due to their ability to generate deliberately false information that can lead to discrediting a person, and to the destruction of infrastructure necessary for human life, and to the deformation of personal and mass perception of the surrounding reality. All this is a consequence of the possible malicious uncontrolled and purposeful introduction of distorted, unreliable, fictitious information into various databases.

4. The development of AI and neural networks and their penetration into all spheres of life indicates the creation of a global information market. As practice shows, the legal systems of various countries are not yet ready to take an active coordinated part in this market and resist the possible negative consequences of its development in the direction of distortion of reality and reality. Some countries (USA) seek to consolidate their market dominance through the development and sale of new information technologies, the second (EU) are trying to develop ways to distinguish generated information from real creative information, the third (PRC) recognize the permissibility of fair use of generated information, the fourth (Japan) hope for self-regulation processes. 

Therefore, there are a number of initiatives in the field of legal regulation of neural networks and generative artificial intelligence (for example, the Project "Measures for managing generative artificial intelligence services" 11.04.2023, etc.).

An analysis of the approaches of domestic and foreign (Great Britain, Australia, Canada, China) legal systems has allowed us to establish that the author of a work of science, literature or art is recognized as a citizen whose creative work it was created" (in various modifications), which raises the question 1) of the need to label content generated by AI, supported by many legal systems (RF, USA) and supranational associations (EU) and 2) clarifying the role and place of AI in this process.

At the same time, from the perspective of pedagogy and prospects for the development of education, including higher education, it seems that:

1.           Artificial intelligence assistance in writing and subsequent evaluation (indicating to the student the mistakes he made and recommendations on ways to overcome such mistakes in the future) of essays, term papers, final qualifying papers and other written assignments as a given, while shifting the focus to new formats and forms of assignments during training and taking (passing) tests and exams to check the learning outcomes, so as not to fight academic fraud, but to prevent it by motivating students to study with tasks that better and more clearly reflect the learning goals and skills important for students for future employment.

Earlier, we addressed the question that the computer algorithm used in the domestic Anti-Plagiarism system, for example, should turn from an "enemy" that students want to "bypass" into an assistant with whom they want to cooperate [18. pp. 1-11]. Another participant of the Round Table "Digital University: Education 3.0" also spoke in this sense (URL: https://news.myseldon.com/ru/news/index/281811555 (date of appeal: 07/26/2023)): "The ban on the use of artificial intelligence in writing student papers is the most unreasonable and unrealizable option. It is better to minimize the motivation of students for dishonest behavior within the educational system. Firstly, the learning process should be enjoyable and exciting. Participation in it should be more interesting than non-participation and evasion. Secondly, it is necessary to learn how to test the student's ability to apply knowledge in various contexts, the ability to think and act. And the assessment of this ability should not be tied to the volume of the prepared text." (Emelianova T. V. Digital technologies in education and human values. //Sustainable development of Russia: a legal dimension : a collection of reports of the X Moscow Legal Forum : at 3 p.m. Part 2. – Moscow: Publishing Center of the O.E. Kutafin University (MGUA), 2023. – pp. 221-226 (522 p.) – Text : electronic).

2. Tasks should consist with this approach (paragraph 1) of online video lectures recorded by teachers independently on appropriate digital platforms equipped with the necessary and appropriate services for this, for example, subtitles in several languages, good annotations, etc., on the one hand, and new formats of online interaction such as practical classes (laboratory work), which will allow students to be active in small groups throughout the lesson, staying "in touch" and instantly receiving feedback, for example, in VR/AR laboratories / on new digital educational platforms / in relevant ecosystems, etc., which, unlike the "repositories" used by universities materials (content) and the MOODLE-based grading tool currently do not provide students with the opportunity to collaborate, share material, communicate with a teacher in real time, including using open educational resources, including aggregators of educational materials created by students, where it is possible to draw their own diagram or find materials from other students. This approach has proven itself well in the author's work on the joint use of the "Whiteboard" tools of the Microsoft Office Teams program, in which students were happy to draw diagrams and carry out classifications together as part of the study of various private law institutions in foreign countries, and then save them for further work. Education will have to adapt.

3. Dialectical pedagogy using generative artificial intelligence can provide faster and more individualized learning than was possible in the past, so teachers should teach new skills, including responsible ways of human-machine communication. In fact, current and future educational and vocational training systems should preserve the idea of people as moral, psychological and strategic beings with a unique ability to make holistic judgments [22. P. 1-35].

Thus, we believe that it is required: 1) a discussion in society about the moral and ethical and pragmatically useful components of the work of neural networks, 2) a thorough analysis of the existing legal regulation of neural networks abroad and at the international level to create adequate domestic regulation in developing approaches to the legislation of the Russian Federation and taking into account the need to respect the economic and technological sovereignty of the country and consolidate the moral and ethical guidelines of scientific works 3) development of algorithms for the operation of the Anti-Plagiarism system, which will allow separating the personal contribution of the author of a scientific text from a machine (algorithm, AI), which will probably require changes in the requirements for scientific papers (dissertations, WRC, etc.) [17. pp. 62-104] (Belikova K.M. Current state and prospects education 3.0 in Russia and abroad. // Sustainable development of Russia: a legal dimension : a collection of reports of the X Moscow Legal Forum : at 3 o'clock Part 2. – Moscow : Publishing Center of the O.E. Kutafin University (MGUA), 2023. – pp. 221-226 (522 p.) – Text : electronic).