Russia wants to limit the use of AI in education. What you need to know

A bill is being developed in Russia that will limit the use of artificial intelligence technologies in education. In the works of schoolchildren and students, it will be acceptable to have no more than 30% of the text generated by neural networks. A bill has also been prepared that will oblige the labeling of AI-generated materials. How the new technology is planned to be regulated and whether it will affect the quality of education is described in the Izvestia article.

What bill is being prepared in the Federation Council

• The Federation Council has begun drafting a bill that will regulate the use of artificial intelligence in the educational process and impose restrictions on its use. The draft law is being developed by a working group at the Federation Council on the legal regulation of various aspects related to AI. The bill has not yet been drafted, and its text has not been included in the legislative framework. However, Natalia Kaspersky, Chairman of the Board of the Association of Software Developers (ARPP) "Domestic Software", revealed the main parameters of the future initiative.

• According to her, the bill is aimed at creating specific rules for the use of neural networks by schoolchildren and students. They will have to disclose the share of the generated information in their works. Such a share cannot exceed 30% of the completed task volume. Teachers will also be given the opportunity to check materials according to clear criteria that will determine whether plagiarism or excessive use of AI has been allowed in their work.

By themselves with the codes: the United States begins to fight against the AI technologies of Russia and other countries

Why is the State Department creating a bureau to counter AI solutions from unfriendly states?

• The draft law aims to develop students' self-learning and critical thinking skills so that they do not rely entirely on the work of AI models and do not delegate tasks to them, passing them off as their own. According to Kaspersky, a quarter of students use neural networks when writing papers, and 87% resort to their help in one form or another. She also called for the bill to prevent the "harmful use" of AI in the social sphere and public administration.

How is AI regulated in Russia now

• At the moment, AI in Russia is regulated at the experimental legal level without a well-established regulatory framework. The Ministry of Finance is trying to fill the gap by proposing its own framework bill "On the basics of state regulation of the application of AI technologies." It is still under discussion and has not been submitted to the State Duma, but it is expected that the law will enter into force on September 1, 2027.

• The draft law of the Ministry of Finance defines a number of terms used in the field of AI. Thus, the agency defines artificial intelligence itself as a set of technologies that mimic human cognitive functions, including self-learning and finding solutions without a specific algorithm, and produce results comparable to what human intellectual activity is capable of.

• The document also establishes that the state supports the implementation of sovereign and national AI models. These include those that are developed and trained in Russia, for which internally generated datasets are used and which are managed by Russian citizens and legal entities. If such AI models meet certain security requirements and process data only on the territory of the Russian Federation, then they can enter the register of trusted AI models. They can be used in government information systems and critical infrastructure facilities.

• The draft law imposes a number of conditions on developers and operators of AI models. Neural networks are required to ensure that their behavior does not lead to discrimination, is not used for prohibited purposes, does not exploit human vulnerability, and stops work in the event of a threat to human health and the security of the state.

• If the service is used by more than 500 thousand people per day, then its owner will have to store the data and content of user messages in order to provide them at the request of law enforcement agencies. He is also required to provide labeling so that the user knows that the generated materials were produced using AI. Separately, the draft law states that AI models should conform to traditional Russian spiritual and moral values, which should be monitored by a special expert body designated by the government of the Russian Federation.

What impact does AI have on education

• When trying to regulate the use of artificial intelligence, a number of features of this technology should be taken into account. First of all, it should be understood that AI models are only an information processing tool and do not fully do the work for the customer. In this respect, they can be compared to calculators, which cannot give the correct result if you do not know how to use them.

• Even if you order a fully completed work from an AI service, it will require careful proofreading and rechecking, and for this you already need to dive into the essence of the problem and study the subject yourself. In order to complete their homework with the help of AI, students and schoolchildren should know how to write scripts correctly, check the facts provided, indicate sources, and edit the resulting text. These actions will ultimately have a positive impact on the educational process and provide new knowledge. Mindless use of AI, which leads to an incorrectly completed task, is equivalent to blind cheating without mastering the topic being studied.

Trust and verify: how neural networks will start treating Russians

Mass retraining of doctors will begin soon.

• When regulating the impact of AI on the educational process, it should also be taken into account that it is impossible to detect generated content with 100% accuracy. Existing AI text detectors themselves show unstable results and can mistake hand-written text created before the advent of neural networks for the product of large language models. Any AI service is trained on data that was once created by humans without the help of modern technologies, and only reflects them in a slightly modified form. They do not create fundamentally new information that a student or student could pass off as their own, but only serve as a high-tech intermediary between the source and the user.

• In order to eliminate the possibility that the use of neural networks hinders education and the formation of critical and logical thinking, it is necessary to rethink the very essence of learning and move from control over the results to control over the process. So, AI will not be able to influence in any way the oral knowledge test based on the results of the work done. If a student manages to defend his work written using an AI model, then this will indicate his immersion in the topic. It is also possible to introduce classroom work into the educational process in conditions where access to AI tools is limited, or to individualize tasks for which AI will not have relevant data.

• A number of negative aspects can be identified in the constant use of AI. It can lead to cliched responses, create the illusion of competence for the user, reduce writing and text construction skills, and lead to neglect of authorship and sources. However, the same problems could haunt students even before the introduction of neural networks if they took an irresponsible approach to their studies and pursued only high grades, rather than gaining knowledge.