Astana, October 16, 2025 —The Think Tank Forum 2025, convened on October 15–16 in Astana by the Kazakhstan Institute for Strategic Studies (KazISS) under the auspices of Astana International Forum (AIF), brought together policymakers, scholars, and thought leaders from around the world to explore ways to rebuild trust in global cooperation under the banner “From Polarization to Partnership: Rebuilding Trust in the International System.” The event opened with welcoming remarks by Magzum Mirzagaliyev, advisor to the President of Kazakhstan, and Zhandos Shaimardanov, Director of KazISS.

Around 40 foreign experts from 20 countries, including heads of leading think tanks, academics, and policy professionals, joined discussions across nine sessions covering topics such as global governance reform, nuclear disarmament, AI ethics, and the transformation of the international security system.

As part of the Forum, a high-level panel titled “AI Without Borders: Ethical Dilemmas and Challenges of the Future” explored how artificial intelligence is reshaping societies worldwide while testing the limits of existing governance structures.

Moderated by Saken Zhumashev, Managing Partner at KPMG Caucasus and Central Asia, the discussion examined how the accelerating pace of AI development is outstripping regulatory frameworks and ethical oversight. The panel addressed the risks of geopolitical fragmentation in AI governance and considered how regional partnerships might bridge the growing trust gap between nations.

Panelists included Bayan Konirbayev, Deputy Chair of the Board of Trustees at the Alem.ai International Center for Artificial Intelligence; Ekaterina Smyshlyaeva, Member of the Mazhilis of the Parliament of the Republic of Kazakhstan; Huseyin Atakan Varol, Founding Director of the Institute of Smart Systems and Artificial Intelligence (ISSAI) at Nazarbayev University; Karamjit Gill, Editor-in-Chief of AI & Society and Professor Emeritus at the University of Brighton; and Virginia Dignum, Professor of Responsible Artificial Intelligence and Director of the AI Policy Lab at Umeå University.

In his address, Dr. Huseyin Atakan Varol emphasized the growing importance of regional contributions to the global AI ecosystem and the need for ethical and inclusive innovation. He described artificial intelligence as humanity’s latest “general-purpose technology,” following in the historical lineage of writing, printing, and the internet. Varol argued that generative AI represents the next transformative leap—one that could eventually bring society closer to artificial general intelligence (AGI).

He referred to the race toward AGI as “the new great game of the 21st century,” pointing out that this emerging contest is defined by three interrelated fronts: developing advanced AI models, producing high-performance computing hardware, and ensuring energy security to sustain large-scale computation.

Amid this global competition, Dr. Varol highlighted Kazakhstan’s proactive efforts to position itself as a responsible middle power in the AI arena. Under the leadership of President Kassym-Jomart Tokayev, Kazakhstan is building a comprehensive legal and technical foundation for AI development, investing in supercomputing capacity, and expanding its energy infrastructure.

Dr. Varol also outlined ISSAI’s key milestones, including the creation of the Kazakh Large Language Model (KazLLM) in December 2024 — the country’s first national generative AI model designed to promote linguistic inclusion and cultural alignment. He further introduced ISSAI’s suite of applied AI systems such as Oylan, Mangisoz, Tilsync, and Beynele, all of which demonstrate Kazakhstan’s growing capability in the field.

Concluding his remarks, Dr. Varol underscored that building public trust in AI requires not only technological advancement but also ethical integration across education, industry, and governance. He called for collaborative action among academic institutions, private enterprises, and policymakers to ensure that AI serves as a force for shared prosperity and human development.

The Think Tank Forum 2025 reaffirmed Kazakhstan’s emergence as a hub for dialogue on technology, ethics, and international cooperation — bridging global perspectives and local innovation under the shared vision of rebuilding trust in the international system.

The full text of Dr. Varol’s speech is provided below:

“Mr. Chair, distinguished panelists, and esteemed members of the audience,

It is my distinct pleasure to participate in the superbly organized Astana Think Tank Forum today. I hope to add my perspective as an AI technologist from the region to our discussion.

We usually tell the history of humankind through important events, such as battles or treaties; or influential people, like generals or leaders; or through ideologies, such as communism or capitalism. However, there is another way. We can describe human history as a progression of general-purpose technologies. These are paradigm-shifting inventions that affect the entire economy and have significant spillover effects. Among these, I find information-related general-purpose technologies to be the most impactful ones. These include the invention of speech, writing, printing press, telegraph, transistor, personal computer, the internet, and, most recently, generative artificial intelligence.

Indeed, generative AI is especially important because many experts predict it is the gateway to artificial general intelligence, a level where AI can perform all tasks at or better than a human. Many postulate that this will lead to a technological singularity, an accelerated timeline of uncontrollable technological development.

Therefore, a huge race is underway to reach artificial general intelligence first. Indeed, we can call the quest to achieve globally deployed artificial general intelligence the new “great game” of the 21st century. AI superpowers, the United States and China, are the two countries leading this race. And this game is played on three levels. The first level is to create the best generative AI models, such as Google’s Gemini, OpenAI’s GPT-5, or DeepSeek’s V3. The second level is to create the computational hardware to train and deploy these AI models. This is what has made Nvidia the most valuable company in the world by market capitalization. The third level is to establish the electrical power generation capacity to run the energy-hungry AI hardware. The next few years will reveal the ultimate winner of this game.

Kazakhstan, under the leadership of His Excellency President Kassym-Jomart Tokayev, is gaining recognition as a responsible, reliable, and respected middle power, and it is also taking part in the generative AI age. The country is establishing a legal framework for AI and has created the Ministry of Artificial Intelligence to concert its efforts. The country is also developing its own generative AI models through fine-tuning of open-source models. It is building data centers and establishing supercomputing clusters. Furthermore, Kazakhstan is upgrading its power grid and commencing the construction of its nuclear power plants. 

Here comes an important question: Why should middle powers like Kazakhstan create their own generative AI models? After all, models capable of high-level performance can be accessed with a subscription. The answer is multifold. Firstly, global AI models such as ChatGPT achieve their best performance in the languages of their primary markets, such as English. Their performance for other languages, such as Kazakh, Thai, or Urdu, is substantially lower. Secondly, they are not aligned for local cultures. What is acceptable in a North American context might not be suitable for another culture. Thirdly, using closed-source services, such as Google Gemini, creates digital dependency and raises data privacy issues. Moreover, these models cannot be easily customized for specific tasks.

Now that we have established the need for creating localized AI models, the next question is, how?

Here, I can share our experience at the Institute of Smart Systems and Artificial Intelligence (ISSAI) at Nazarbayev University in creating the Kazakh Large Language Model (KazLLM). We founded this institute in 2019, when AI was not yet a headline topic. When we were asked to develop the KazLLM in early 2024, we were ready. We had already established our institute as the largest repository of open-source deep learning models, datasets, and training recipes in Central Asia and had also developed a strong local talent pool.

In the KazLLM project, our main challenge was creating the dataset, as the amount of available data in the Kazakh language is much lower compared to English. Another challenge was learning how to use supercomputing clusters for efficient and effective model training. Kazakhstan’s greatest strength lies in its youth. The young researchers at ISSAI tackled all of these problems, and we delivered the KazLLM in December 2024 to Astana Hub.

However, this was not an end but a beginning for us. Since then, our institute has introduced a long list of AI models and services. These include Oylan, a multimodal language, audio, and vision model capable of operating in Kazakh, Russian and English; Mangisoz, a text and speech translation engine; Tilsync, a real-time subtitling service; and Beynele, a culturally-tailored image generation engine. Essentially, when you empower young talent with computational resources in a conducive environment of a research university (Nazarbayev University in this case), magic happens.

I would like to conclude my talk with a note of cautious optimism. Kazakhstan has the potential to be an important player in the AI age, especially if it focuses on niche areas where it has inherent strengths and pressing needs. However, creating AI models and establishing infrastructure is only one part of the challenge. The integration of these models into processes across business, industry, government, education, and healthcare to improve the quality of life and safety of citizens, and to increase economic output, is the real major issue. This can only be achieved through close collaboration between academia, the state, and industry, all operating under a legal framework for responsible AI that draws from the best practices around the world.”