National AI Capacities Across Central Asia: What the GIRAI 2023 Assessment Found
The Global Index on Responsible AI (GIRAI) 1st Edition evaluated national AI capacities across three sub-dimensions: institutions, investments, and competencies. Across Central Asia, the National Responsible AI Capacities dimension is where the regional differences are sharpest — in institutional depth, workforce pipeline, commercial AI activity, and the connection (or disconnection) between capacity and governance. The full GIRAI data is at global-index.ai. This post reads the dimension comparatively across all five Central Asian states.
The Region at a Glance
Institutions: Who Is Doing AI Research
Kazakhstan has the most institutionalized AI research base in the region. According to the Ministry of Science and Higher Education of Kazakhstan, 24 universities and research centers are engaged in AI research or development to varying degrees. The Concept for the Development of Artificial Intelligence for 2024–2029 identifies four institutions with advanced computing infrastructure as the current backbone: Eurasian National University (L.N. Gumilyev), Al-Farabi Kazakh National University, Satpayev University, and Nazarbayev University. These four anchor the national AI research capacity in a way that has no equivalent elsewhere in the region.
The limitation the GIRAI assessment identifies is not in the number of institutions but in the governance of their direction. AI research in Kazakhstan is heavily oriented toward state priorities — healthcare AI, manufacturing quality control, government services automation — and the mechanisms by which civil society, private sector, or independent academic voices feed into the national AI agenda are constrained. The two-week public consultation window for the AI Concept is the most visible symptom of a process that formally exists but is functionally narrow.
Kyrgyzstan’s institutional AI capacity runs through language. The National Commission for State Language and Language Policy under the President is the named implementing agency for both key AI language policy documents — the Cabinet Action Plan and the State Language Policy Programme — giving national AI capacity a specific institutional home outside the typical tech ministry structure. Academic institutions engaged on AI in Kyrgyzstan include those working on NLP for the Kyrgyz language, and civil society organizations have also been found active in the AI space. The GIRAI assessment captures this institutional landscape at the point when the Digital Code’s Chapter 23 had been drafted but not adopted — meaning the governance framework for AI institutions was in preparation but not yet in force.
Tajikistan’s institutional story is dominated by a single actor: zypl.ai, the AI company founded by AI Council chair Azizjon Azimi. By the time of the GIRAI assessment, zypl.ai was providing AI underwriting for one quarter of all loans across eight financial institutions — a market penetration that makes it, by some measures, the most commercially consequential AI deployment in the region at the time of the assessment. The AI Council under the Ministry of Innovative Development serves as the government’s AI institutional anchor. The strategy targets seven higher education institutions with AI departments by 2026 and 5,000 qualified AI specialists by 2040. What is notable is that the target does not include responsible AI or ethics qualifications — the capacity being built is technical, not governance-oriented.
Uzbekistan’s AI institutional landscape is harder to read from open sources than any other country in the region. The Advisory Council on Artificial Intelligence, established in August 2023 under the Ministry of Innovative Development, is the most recent institutional addition. IT Park — Uzbekistan’s technology zone established to attract IT companies and startups — operates as a parallel institutional anchor for the private sector. The USAID Digital Ecosystem Country Assessment provides the clearest available picture of what the ecosystem looks like in practice. The GIRAI assessment found limited documentation of civil society or academic participation in policy dialogue — not because they are absent, but because data on non-government engagement was difficult to obtain.
Turkmenistan has university AI programme components, confirmed by the GIRAI assessment. What is absent is the full institutional architecture that would give those programmes national consequence: no academic publishing on AI, no expert participation in regional or international forums, no research institution with the infrastructure or mandate to translate AI work into policy dialogue. The components exist in isolation from each other and from any governance process.
Investments: Where the Money Is Coming From
Across Central Asia, AI investment is primarily state-driven. Private sector and international donor investment is present in each country but plays a secondary role to government commitment — which means the direction of AI capacity is shaped more by state priorities than by market demand or civil society input.
Kazakhstan’s investment model is the most explicitly articulated. The 2023 Concept on Digital Transformation, Development of ICT, and Cybersecurity Industry commits to creating a national AI platform by December 2024, with a roadmap for AI development completed by December 2023. The National Development Plan until 2029 — with its headline target of GDP doubling to $450 billion — names AI as a tool for healthcare diagnostics specifically. The investment is real and the timelines are specific. The gap the GIRAI researcher identifies is between the scale of the commitment and the thinness of the governance framework surrounding it: frameworks for algorithmic accountability, human oversight, and transparency are underdeveloped relative to the pace of implementation commitments.
Tajikistan’s investment model is multi-source: state budget, development partners, and private sector. The zypl.ai commercial investment is self-sustaining — the company generates revenue from its financial services AI — which makes Tajikistan unusual in having a commercial AI actor that is not dependent on state subsidy. The broader strategy targets AI-related activity at 5% of GDP by 2040 and 1% by 2026; the investment pathway to those targets is not yet fully specified.
Uzbekistan’s investment in AI flows primarily through the 2021 presidential programme’s identified pilot sectors — finance, banking, tax, e-government — and the IT Park ecosystem that supports private sector technology activity. The World Bank’s November 2023 Country Climate and Development Report frames a complicating context: rapid population and economic growth will drive significant increases in emissions, placing strain on natural resources. The energy implications of AI infrastructure investment have not yet entered the responsible AI conversation in Uzbekistan.
Kyrgyzstan’s AI investment is smaller in absolute scale than Kazakhstan’s or Uzbekistan’s, but notable for its specificity: the binding national language policy documents commit government resources specifically to AI for the Kyrgyz language, which means the investment is oriented toward a rights-adjacent outcome (linguistic preservation and inclusion) rather than purely toward economic efficiency. This is unusual in the regional context.
Competencies: Skills, Engagement, and the Ethics Gap
The National Responsible AI Capacities dimension asks not only whether a country has AI research institutions and investment, but whether the people and organizations responsible for AI governance have the competencies to govern it responsibly. Across Central Asia, this is where the assessment finds the most consistent weakness.
Kazakhstan’s AI Concept acknowledges the competency gap at the baseline: «many citizens lack understanding of how these technologies function across sectors.» This is a candid admission in a strategy document — and one that sits awkwardly alongside the two-week public consultation window for that same document. The GIRAI researcher flags this as a question about genuine public engagement rather than formal compliance: a consultation designed to satisfy a procedural requirement, not to draw on citizen knowledge that the document itself acknowledges is currently insufficient.
Tajikistan’s AI Council has acknowledged, in conversations with the GIRAI researcher, that members of parliament lack the knowledge required to formulate AI standards or regulatory policy. The gap is not just between strategy documents and practice; it runs through the institutions responsible for closing it. The AI Council’s own composition — centred on Azimi and zypl.ai — means that the entity responsible for AI governance is also the country’s primary commercial AI actor. The conflict between development promotion and accountability oversight is unresolved.
Uzbekistan’s competency picture is the hardest to read. The GIRAI researcher found low involvement of civil society organizations and academic institutions in AI policy dialogue — their engagement identified as crucial for fostering innovation, providing expert insight, and ensuring ethical development, but mechanisms to integrate them into the national AI strategy were not yet in place during the research period. Private sector engagement was similarly difficult to document: data on private sector AI activities was sparse, making it hard to assess the full extent of non-government engagement.
Kyrgyzstan’s competency story runs through the organizations that engaged with the GIRAI assessment process itself: the academic institutions working on AI and NLP for the national language, the civil society actors covering algorithm effects in media and economic reporting, and the government agencies implementing binding policy commitments on AI in language and education. These are not AI governance specialists in the narrow sense; they are actors with specific knowledge of specific domains where AI intersects with their work. That distributed competency is what the GIRAI assessment captured and confirmed.
What the Dimension Shows
The National Responsible AI Capacities dimension is where the regional differences are most quantifiable — in number of research institutions, in scale of government commitment, in commercial AI deployment. It is also the dimension where the relationship between capacity and responsible governance is most fraught. A country can have twenty-four AI research institutions and a two-week public consultation window. A country can have a commercially dominant AI loan-underwriting firm and no accountability framework for algorithmic credit decisions. The GIRAI assessment captures both the capacity and the gap. The question subsequent assessments will be able to answer is whether they move together.
Based on the GIRAI 1st Edition (2023) National Responsible AI Capacities dimension findings for the five Central Asian states. Data source: global-index.ai. Regional hub: IDFI (Georgia). Kyrgyzstan country researcher: Aziz Soltobaev. This is a regional thematic comparison by KG Labs.
