In the networks of Themis: what awaits the mystery of Russian users' dialogues

The interference of artificial intelligence in personal and business life has become a global trend, but the topic of the safety and secrecy of dialogues with neural networks is only beginning to receive due reflection in the legislative field. The news that OpenAI may disclose ChatGPT user logs at a court request caused There is a wide response in Russia. Izvestia discussed with parliamentarians and data protection experts the risks of personal information disclosure, regulatory gaps, and the future of Russia's digital sovereignty. The Center for Legal Assistance to Citizens in the Digital Environment (under Roskomnadzor) also expressed its position.

The secret of communicating with AI

In a recent interview, Sam Altman, head of OpenAI, said: upon receiving a court request, the company will be required to transmit logs of users' conversations with the neural network, despite Altman's own privacy policy and personal position against such practices. The company that created the symbol of a new era, the ChatGPT neural network, explains that the decision is due to the lack of clear international and national standards for ensuring the secrecy of such data. Currently, OpenAI stores deleted conversations for 30 days and can use the correspondence to improve the service or identify violations.

Photo: IZVESTIA/Dmitry Korotaev

In Russia, as the national Digital code is being prepared, there is also a problem with data privacy in AI. Domestic experts note that the situation does not differ from global realities, which means that it will be possible to learn from the experience of legal regulation from other sectors of the digital world.

Building a new AI world: Russia begins another digital breakthrough

But the legislative framework and human resources are not keeping pace with the dynamics of innovation.

Alexander Babakov, Vice Speaker of the State Duma and head of the inter-factional working group on the development of laws on the use of artificial intelligence, emphasizes the importance of deep regulation for the operation of neural networks in the national interest.

— If we want to be a sovereign state, then both AI technologies and tools should be a priority in the development of neural networks, as well as the regulation of the topics discussed, taking into account the awareness of the threats they may pose. Not everything can be left to the so—called market," the expert believes.

According to the Vice Speaker, it is important that the process of implementing neural networks and accessing their data does not become uncontrolled. This means that there should be a fairly clear legal regulation, which does not exclude, but presupposes moving forward.

Photo: IZVESTIA/Eduard Kornienko

Senator Artyom Sheikin confirms that the issue of access by courts or law enforcement agencies to user dialogues with neural networks remains largely debatable and unresolved at the legislative level.

According to him, the user often voluntarily transmits information to the service. The service can use this data both for learning algorithms and for other purposes, which increases the risk of leaks or unauthorized access. As a result, confidentiality guarantees depend on procedural nuances.

Data disclosure practices

Experts from the Center for Legal Assistance to Citizens in the Digital Environment (CCI) at Roskomnadzor state that there is no direct legal norm in Russia governing access to correspondence with neural networks. However, the general provisions of the laws allow law enforcement agencies and courts to request such information to solve problems. The law "On Operational Investigative Activities" and other regulations contain mechanisms for such requests, including cases related to security and criminal cases.

There has been no official information yet that Russian law enforcement agencies have already contacted foreign IT companies for AI correspondence logs, the Central Control Center told Izvestia.

Photo: IZVESTIA/Sergey Konkov

The center believes that dialogues with the neural network, including metadata, should be interpreted as personal data, since they are associated with a specific user, characterize his actions and vocabulary. Service owners usually directly receive the right to use this data, including distribution. Companies often form their rules taking into account national legislation, so the user should be aware of the possible risks when transferring sensitive information to neural networks, they say in the CCI.

Playing by the rules: Russia has developed a code of ethics for working with AI for financial organizations

The document is designed to ensure the safety of technology application in the market

Nver Gasparyan, Vice President of the Federal Chamber of Lawyers of the Russian Federation, notes that law enforcement officers may request access to citizens' dialogues with neural networks, but the procedural status of this information has yet to be consolidated. Such information should be treated as confidential and require a court order. Otherwise, there are no guarantees of protection, he points out.

As explained by a leading researcher at the Laboratory of Legal Regulation of the Institute of Legislation and Comparative Law (IZAK) Pavel Kabytkov, the right to privacy of correspondence, telephone conversations and other communications is guaranteed by Article 23 of the Constitution, but this right is not absolute.

Photo: IZVESTIA/Dmitry Korotaev

— It may be limited, for example, in the investigation of crimes or the fight against terrorism. Even special neural network agreements often restrict the transfer of personal data to third parties, and processing by the operator is subject to general personal data legislation," the expert explains.

The position of experts on cybersecurity and law

Vladimir Vinogradov, Dean of the HSE Faculty of Law, notes that currently the issues of transferring AI correspondence logs to the court are not directly regulated, and corporate service policies may allow such disclosure upon request from law enforcement agencies.

"Privacy and correspondence are protected by the Constitution,— Vinogradov explains. — A violation is allowed only on the basis of a court decision, for example, during the investigation of a criminal case or the fight against terrorism. However, due to the absence of a separate law on AI, access by courts and law enforcement officers to correspondence with neural networks can be implemented in different ways - either as an interaction with a legal entity (service), or as a public search query.

Photo: IZVESTIA/Anna Selina

Alexander Ponomarev, product director of a domestic company, is convinced that Altman's statement is not a feature of the United States, but a global reality that will persist until special guarantees for AI dialogues appear.

— In Russia, there are rules on the secrecy of messages, but for services recognized as organizers of information dissemination, there is a duty to store and provide data to authorities in accordance with the SORM rules. This also covers chat logs. For foreign platforms, however, everything depends on cross—border jurisdiction and data acquisition is extremely slow or even impossible," the expert states.

Transition to faces: authorities will start blocking fraudulent content created by AI

Will the new measures help solve the problem with deepfakes?

The CCI experts state that "the mechanisms of interaction between business and the state should be reflected in future laws that are currently being worked on, primarily in the Digital Code. It is necessary not only to develop technologies, but also to protect citizens from the threat of leakage of personal information, they believe.