Russian Federation

We are convinced that the very launch of in-depth and depoliticized discussions on all agenda items established by UNGA Resolution 79/325, with equal participation of all States, would constitute a successful outcome of the Global Dialogue. We are hopeful that the first Global Dialogue will be held in an atmosphere of equal and mutually respectful discussions and result in a comprehensive, impartial and non-prescriptive Co-Chairs' summary reflecting the full diversity of approaches to AI governance and the interests of all Member States, in particular the needs of developing and least developed countries. This would contribute to bridging the digital divide and achieving the Sustainable Development Goals. Considering the Global Dialogue in a longer-term perspective, we would also regard the following as success: Inclusion of "high-risk" trends in the mandate. Generative AI and autonomous AI agents should become cross-cutting scenarios for consideration within the framework of safety and accountability; otherwise, the Dialogue will lag behind technological reality. Inclusive list of initiatives. Creating a registry of international, regional and intergovernmental initiatives, including platforms of the African Union, ASEAN, BRICS and other mechanisms of developing countries, not only those established by Western countries. Tasking the Scientific Panel with taking stock of regulatory models and ethical codes beyond North America and Western Europe. Roadmap and measurable outcomes. Agreeing on a phased plan with specific measurable utcomes: analytical reports, model guidelines, open solutions catalogues, mechanisms for validation of cultural and linguistic diversity. Stipulating that the roadmap does not duplicate other UN formats, and that working formats (e.g. working groups) are formed on the principle of fair geographical representation. Thus, clear and understandable steps and action plans will transform the first session from a declaratory event into a working mechanism for coordinating efforts towards fair and safe AI development.

• Safe, secure and trustworthy AI
• AI capacity-building
• Transparency, accountability, and human oversight
• Open-source software, open data and open AI models

Safe, secure and trustworthy AI as well as transparency, accountability and human oversight form a unified framework. The rapid development of AI opens up new opportunities but also generates new risks. We consider it necessary to develop and deploy AI as a safe and trustworthy technology, ensuring risk minimisation and effective control in accordance with domestic and international law. Promotion of safe, secure and trustworthy AI is fundamental priority. It is closely intertwined with national security, protection of fundamental rights and freedoms, strengthening trust in AI technologies and accelerating digital transformation while preserving digital sovereignty. Human oversight remains important for maintaining control over key decisions, preventing harmful consequences and ensuring AI systems functioning in accordance with legal and ethical standards. As AI is deployed in critical sectors and public administration systems, risks of opaque decisions, errors and abuse increase. Without reliable mechanisms of interpretability, audit and human control, it is impossible to ensure legitimacy and sustainability of such systems. AI capacity-building is a necessary condition for equitable global development. Now, the gap in capabilities between developed and developing States (e.g. infrastructure, digital resources, skilled personnel) is enormous. The potential of AI technologies cannot be fully implemented without addressing this problem. We seek to prevent digital colonization trends, whereby least developed countries would be forced into full dependence on global leaders. Open-source software, open data and AI models development is key for digital transformation and bridging the technological divide. Open ecosystems enable all countries to build competencies, adapt solutions and participate in global value chains. Closed-source AI models complicate identifying vulnerabilities and assess compliance with information security requirements. This allows malicious backdoors, which may lead to data leaks or deliberate unauthorized access.

Taking into account the dynamics of technological progress and identified imbalances in international initiatives, it would be advisable to refine and supplement the agenda in a targeted manner. First, the topic of generative AI is not sufficiently distinct to be highlighted as a standalone topic, but it is important to anchor it as a cross-cutting theme within existing priorities, primarily "safe, secure and trustworthy AI", "transparency and accountability", and "social and cultural implications". This is due to the fact that generative AI simultaneously affects issues of trust in information, manipulative practices, transformation of labour markets and cultural industries. At the same time, singling it out as a separate topic may lead to fragmentation of the agenda and rapid loss of relevance given the fast pace of technological evolution. Second, it would be appropriate to add a new theme: "Ensuring equitable access to AI computational infrastructure and resources (compute power)". This issue is critically important as access to compute, rather than only data or algorithms, is becoming a key factor of global inequality in AI. Without addressing it, efforts to develop open models and capacity-building risk remaining limited. Third, it is recommended to expand the existing item on open data into a related theme: "Data governance and cross-border data flows", including issues of quality, representativeness and data sovereignty. Taken together, these aspects would enhance the practical applicability of the agenda, align it with real technological constraints and strengthen the inclusive nature of global AI governance. On the national level, effective AI governance requires adequate regulations, i.e. legislation and strategies. Nowadays, many countries are still elaborating such documents. It could be mutually beneficial to exchange information and experience on AI regulations and best practices on their implementation through UN bodies.

Simplifying access to AI resources is inseparable from the objectives of safe and trustworthy AI. The growing concentration of AI capabilities in a narrow group of countries and companies raises concerns regarding unequal access to this technology. Only equitable access to computational resources, data and foundational models can ensure the security and sovereignty of States. Such access should not be accompanied by politically motivated restrictions nor used as an instrument of pressure or covert control over data and infrastructure. The widespread use of foreign AI models may create risks for national security and technological sovereignty. The main challenges stem from the fact that such models may be trained on datasets containing unacceptable narratives (e.g. Russophobia, falsification of history, fake news). We are concerned with attempts by certain States to aggressively impose their AI solutions and standards, apply domestic legislation exterritorialy and restrict the AI potential of other countries through trade restrictions and export control measures. This undermines the technological sovereignty of developing countries. It would be unacceptable to try elaborate the international system of AI governance based on politically biased forums and initiatives with a non-transparent list of participants.

We expect the Dialogue to become a neutral platform for discussing international cooperation, exchanging best practices and promoting open, transparent and inclusive discussions on AI governance in full accordance with UNGA Resolution 79/325 (i.a. excluding discussions on AI in the military domain). It should become a catalyst of a genuinely equitable international AI governance system, with the UN playing a central and coordinating role. At the international level, the Global Dialogue can and should contribute to: - facilitating the bridging of digital divide by supporting developing States in strengthening technological sovereignty within efforts to achieve the 2030 Agenda for Sustainable Development; - shaping an international AI governance system based on respect for sovereignty and ensuring national security; - promoting principles that ensure equal rights and opportunities of nations in AI technologies development and use (including principles of safety and trustworthiness, transparency, technological sovereignty, security and reliability of source data, integrity of the innovation cycle, most efficient use of AI technologies, support for competition, openness and accessibility, continuity, protection of human rights and freedoms, preservation of linguistic and cultural diversity); - establishing cooperation mechanisms with interested countries in developing trustworthy AI systems; - forming repositories of tools for control and ensuring safety and trustworthiness of AI technologies, trustworthy AI solutions and models, as well as a resource for exchanging AI models and open-source software solutions, taking into account national legislation. The Global Dialogue should be conducted in a depoliticized manner. Its ultimate goal should be contributing to the achievement of the Sustainable Development Goals and to enhance practical cooperation among Member States.

The Global Dialogue would benefit from close cooperation with relevant UN bodies, in particular the ITU. The Union could share information about its work on developing technical standards in AI, including those related to safe, secure and trustworthy AI systems. The Global Dialogue could also benefit from information exchange with related UN processes, namely the Global Mechanism on ICTs in the Context of International Security and the Ad Hoc Committee to Elaborate a Comprehensive International Convention to Combat the Use of ICTs for Criminal Purposes. The Global Dialogue should also build upon inclusive intergovernmental formats and initiatives reflecting the views of a wide range of States, including developing and least developed countries, and maintain engagement with them. Particular attention should be given to platforms with wide participation where AI governance approaches are developed on the basis of equality, respect for sovereignty and the imperative to bridge digital divides. UNESCO's experience in the field of AI ethics should be taken into account, particularly the 2021 Recommendation on the Ethics of Artificial Intelligence. Relevant examples include work carried out within the Shanghai Cooperation Organisation (SCO), BRICS, the Commonwealth of Independent States (CIS) and the Collective Security Treaty Organization (CSTO). Among these are: - the SCO Heads of State Council Statement on further deepening inter-State cooperation in AI (Tianjin, 2025); - the BRICS Leaders' Statement on global AI governance (Rio de Janeiro, 2025); - the CIS Heads of State Statement on AI cooperation (Moscow, 2024); - the CSTO Parliamentary Assembly Council Statement on AI technologies (Almaty, 2024). Aligning the Dialogue with such initiatives would help avoid narrow or fragmented discussions, ensure broader consideration of national and regional approaches, and strengthen practical cooperation among governments.

Since States bear responsibility for the development and implementation of AI strategies, policies and regulatory frameworks, we believe that practice-oriented contributions by governments will ensure genuinely open and multifaceted discussions on AI governance, as well as opportunities for knowledge exchange and mutual learning. We believe that other stakeholders (primarily business and academia) can contribute to the Global Dialogue by providing objective and depoliticized technical data and research. However, this should imply adequate representation of voices from all countries, especially developing and least developed ones. Although academia and business circles offer a unique perspective and valuable insight into the specifics of AI development, application and regulation, their contribution must be limited with a consultative and supportive role. The Global Dialogue should not compete with existing initiatives, but to connect them into a politically legitimate and inclusive architecture, integrating different approaches and interests. For this purpose, the Dialogue should rely on authoritative platforms reflecting the diversity of governance models, i.e. BRICS, SCO, ASEAN, African Union and others, promoting sovereignty, development and equitable access. Its added value lies in its ability to serve as a connecting link between different models, bring them together on a universal platform and thereby enhance the legitimacy of existing approaches, reducing their perception as unilateral.

The UNGA Resolution 79/325 already establishes the Global Dialogue as a platform open to all governments and stakeholders. It is essential to ensure that this principle is respected. At the same time, the Dialogue should learn from the experience of the Internet Governance Forum (IGF) where excessive diversification of participants and thematic dilution led to the blurring of its original mandate, i.e. Internet governance. Prevalence of human rights rhetoric, in defiance of regulatory, technical and technological discussions, has prevented it from developing in practical, scientific and normative dimensions. This trajectory should be avoided for AI. Currently, discussions in many AI partnerships and initiatives are dominated by technologically leading States, large companies and affiliated expert communities, while several important voices remain insufficiently represented. The latter include: - countries of the Global South and "second-tier" economies whose participation in elaborating norms is limited; - engineering and scientific communities outside Big Tech, including compute infrastructure and fundamental research specialists; - sectorial experts (medicine, biotechnology, chemistry, energy etc.) where AI already directly influences critical processes; - non-English linguistic and cultural communities as well as policymakers from developing countries facing the issues of integrating AI; - representatives of the education sector due to the influence of AI technologies on education. An excessive focus on commercial interests and Western interpretation of human rights risks undermining the scientific and technical foundation and universality of the discussion, which is inadmissible for a platform which claims for global legitimacy and practical effectiveness. The Global Dialogue should instead become a balanced platform integrating humanitarian, technical and regulatory dimensions.

At the current stage, without an agreed roadmap, it is difficult to define formats precisely. However, two promising directions based on nationally tested practices can already be identified: International regulatory sandboxes A mechanism under which interested countries voluntarily test specific AI governance elements (regulatory models, certification approaches, audit mechanisms) in controlled but real environments. A possible linkage with the ITU would allow the transition from pilot results to full standards. The key benefit is in reducing the gap between discussion and implementation. Evidence-based policy-tech interface Joint sessions where policymakers, engineers and scientists work together not via general discussions but by analysing model architectures, data limitations, compute constraints and real compromises. Such discussions would navigate the conversation away from mottos to concrete technical and organizational solutions. This format can be strengthened with a scientific component ("ask a scientist, not a vendor"), with demonstrations from applied and fundamental fields. Scientists would demonstrate where AI is really changing their work, which barriers remain and where systems make mistakes. Policymakers and diplomats would ask questions directly, which would help focus on evidence-based scientific reality and establish a common factual basis for further regulation. This could help shift discussions from abstract rhetoric turning the Global Dialogue into a platform where norms emerge from practice and verified data.

Comprehensive strategic documents play a key role in AI governance. They should enable fast and effective development, deployment, implementation and use of AI systems without creating unnecessary barriers while mitigating potential risks. This is the basic approach in the foundation of all relevant documents in Russia. E.g. the National Strategy for AI Development in the Russian Federation until 2030 provides a holistic framework combining technological development with risk management, human-centered principles and support for innovation, capacity-building, research and integration of AI across various sectors. In addition to the Strategy, in 2021, Russia elaborated a holistic system of ethic norms. Developed within the AI Alliance by over 1000 experts from 42 countries in law, sociology, computer science, education and philosophy and aligned with the UNESCO Recommendations, the Code of ethics in AI establishes standards for safe and ethical development, integration and use of AI technologies by both private and public organizations. A national Commission on the implementation of the Code has been established together with the system of authorized representatives on AI ethics. The AI Alliance serves as a core for consolidating market in providing technological leadership. An important practical instrument is Russia's Federal Law on experimental legal regimes in the sphere of digital innovations (258-FZ) which established regulatory sandboxes. 18 experimental legal regimes have been launched, including for AI. Despite foreign restrictions, Russia remains a noticeable participant of the global technology race as one of rare countries with their own LLMs (GigaChat, Kandinsky, YandexGPT) and prominent experts globally. Internationally, the commitments of SCO, BRICS, CIS and CSTO should be considered. Their political documents mentioned in question 14 cover AI governance. In 2024, being committed to equitable multilateral dialogue, Russia established the International AI Alliance which expanded in 2025 with national associations from BRICS and other countries joining it.