Tajikistan in the GIRAI 2023 Assessment: A Strategy Without a Framework
In September 2022, Tajikistan adopted the Strategy for the Development of Artificial Intelligence for the Period up to 2040 — becoming the first country in Central Asia and the first low-income country in the world to put a dedicated national AI strategy in place. When the Global Index on Responsible AI (GIRAI) 1st Edition assessed Tajikistan the following year, that document was the most significant thing the governance landscape had to show. The full GIRAI scores for Tajikistan are available at global-index.ai. What the assessment found, alongside the strategy, was an almost complete absence of the responsible AI infrastructure that would give it ethical content.
The Three-Dimensional Breakdown
GIRAI structures every country profile across three top-level dimensions evaluated through thematic areas and actor categories. Legend: ✓ = documented evidence approved by GIRAI headquarters · ◐ = drafted, planned, or partially documented · — = no documented evidence at the time of assessment. Full per-indicator scores at global-index.ai.
Dimension 1 — Responsible AI Governance
| Thematic area | Status | Evidence (Tajikistan) |
|---|---|---|
| Enabling policies | ✓ | National AI Strategy 2022–2040 adopted September 2022 — first in Central Asia, first in any low-income country globally. Headline target: 5% of GDP from AI by 2040, 1% by 2026. |
| Rule of law | — | No accountability or human-oversight provisions in the strategy. AI Council members have stated parliament lacks the knowledge to formulate AI standards. |
| Technical standards | — | No national AI technical standards in force. |
| Technology-specific regulation | — | No regulation of AI in finance (despite zypl.ai underwriting 25% of loans across 8 banks), surveillance, healthcare, or other deployed domains. |
Dimension 2 — Human Rights and AI
| Thematic area | Framework | Gov. action | Private sector | Civil society | Academia |
|---|---|---|---|---|---|
| Freedom of Expression | — | — | — | — | — |
| Public Participation | — | — | — | — | — |
| Data Protection | — | — | — | — | — |
| Cultural & Linguistic Diversity | — | — | — | — | — |
| Health & Well-Being | — | — | — | — | — |
| Children’s Rights | — | — | — | — | — |
| Indigenous Data Sovereignty | — | — | — | — | — |
| Bias & Unfair Discrimination | — | — | — | — | — |
| Gender Equality | — | — | — | — | — |
| Education | — | ◐ | — | ◐ | ✓ |
| Environmental Protection | — | ◐ | — | — | — |
| Labour Protection | — | — | — | — | — |
Dimension 3 — National Responsible AI Capacities
| Sub-dimension | Status | Evidence (Tajikistan) |
|---|---|---|
| Institutions | ◐ | AI Council under Ministry of Innovative Development, chaired by Azizjon Azimi. Strategy targets 7 higher-education institutions with AI departments by 2026 and 5,000 AI specialists by 2040 (technical, not governance-oriented). zypl.ai is the dominant private-sector institution — and also the country’s most commercially consequential AI deployment. |
| Investments | ✓ | Multi-source: state budget, development partners, private sector. zypl.ai is self-sustaining via revenue from financial-services AI — unusual in the region for not depending on state subsidy. Pathway to the 5%/1% GDP targets not yet specified. |
| Competencies | — | AI Council has stated parliament lacks the knowledge to formulate AI standards. Governance capacity to oversee deployments is acknowledged as absent. Conflict between development promotion and accountability oversight unresolved (the same Council promotes AI and is led by the country’s primary commercial AI actor). |
The Strategy: Ambition at Scale
The 2022–2040 strategy sets a headline target: AI-related activity to reach 5% of GDP by 2040, with an intermediate milestone of 1% by 2026. The rationale, articulated by Azizjon Azimi — Chairman of Tajikistan’s AI Council and founder of Central Asia’s first AI laboratory — is that AI offers a route to avoid the middle income trap by leapfrogging traditional industrialization stages. Four pillars structure the strategy:
- Smart Regionalization — Safe City surveillance systems, smart meters, and billing infrastructure across Dushanbe and regional cities
- Education and Human Capital — AI departments in 7 higher education institutions by 2026; 5,000 qualified AI specialists trained by 2040
- E-Government — unified state database, a national Open Data portal, AI-integrated public services
- Industrial Acceleration — AI applied to manufacturing quality control, labor productivity, and carbon reduction
Implementation is staged: phases one and two (2023–2025) cover legal and institutional foundations; later phases move toward full sector integration and, ultimately, technology export. The financing model is multi-source — state budget, development partners, private sector.
Some of this is already running. By the time the GIRAI assessment was conducted, zypl.ai — the AI company Azimi founded — was providing AI underwriting for one quarter of all loans across eight financial institutions in Tajikistan. Telecom providers had built in-house churn-prediction models. Career counseling algorithms were being piloted in schools. The technology deployment is real and moving. What it is not accompanied by is governance.
Responsible AI Governance: A Development Agenda, Not an Accountability One
The GIRAI assessment identifies the same gap that Oxford Insights documented in its own Tajikistan spotlight: sector regulators lack AI training, no formal AI-specific regulations exist beyond the strategy document, standardized guidance across agencies is absent, and responsible AI approval standards remain unwritten. The strategy commits to a legal and institutional framework — but stages it for later phases, after the technology is already deployed.
The GIRAI researcher’s assessment is direct: human oversight of AI operations, safety standards, accountability mechanisms, proportionality requirements, transparent procurement, and access to remedy do not appear in the strategy or in related government initiatives. The AI governance framework that exists is a development framework. The political environment reinforces this. Policymaking operates without public scrutiny; the government treats AI as a domain of progress, and critical examination of AI risks is regarded as an impediment rather than a component. The result is a governance conversation that, in the researcher’s words, is «building more walls around AI implementation, protecting it from critical views.»
Human Rights and AI: Education and Environment, Little Else
No national framework or legislation addresses the potential risks of AI for human security, privacy, or civil liberties. Public discourse on the human rights dimension of AI is thin across both state and non-state actors.
The clearest exception is Rustam Gulov, a new media expert writing for CABAR.asia, who tested AI tools for journalistic use and identified structural barriers to adoption in Tajikistan’s media environment: limited internet access, insufficient professional skills, and press freedom constraints. The piece is a practical adoption assessment rather than a rights critique, but it represents one of the few critical engagements with AI’s societal effects published in the country.
Two thematic areas showed positive evidence in the GIRAI assessment. Education: the AI Council ran awareness events at universities — including a November 2021 masterclass titled «Creating Opportunities with Artificial Intelligence» at MSU Tajikistan delivered by Azimi himself, feeding students into zypl.ai’s educational pipeline — and several universities have introduced AI courses into their programs. Environmental protection: the strategy names green economy as a development goal, and AI is included among the tools referenced in the EBRD Green City Action Plan for Dushanbe.
Beyond those two areas, the GIRAI researcher found the human rights dimension largely absent — not suppressed in any specific instance, but simply not present as a subject of policy, advocacy, or academic inquiry.
Surveillance: Reported but Unconfirmed
The country context report includes a finding that sits at the edge of what the GIRAI framework can document. The researcher received accounts — in direct conversations — that Tajikistan’s security services may be using AI for surveillance and social crediting. This cannot be confirmed from open sources without substantial risk to sources.
What is documented: reports from two years before the assessment described Tajikistan’s intent to integrate Chinese facial recognition technologies into the existing «Safe City» network in Dushanbe — a city-wide street camera system already operational at the time. The rollout coincided with a government decree requiring citizens to replace paper passports with biometric IDs, with mobile network access threatened for those who refused. The researcher’s reading is that this was likely intended to build the digital data infrastructure required for AI tracking systems. As of the GIRAI research period, no confirmed open evidence of an operational facial recognition system existed. The strategy’s first pillar — Smart Regionalization — explicitly includes expansion of surveillance systems, which sits alongside these concerns without resolving them.
National AI Capacities: Skills Without Governance Depth
Tajikistan’s capacity-building is concentrated on entry-level skills and popularization: university courses, student awareness events, extracurricular programs. The UNICEF Tajikistan and Ministry of Labour «Professional Guide and Youth Portal» initiative fits this pattern. The AI Council has acknowledged that ethics is part of the eventual agenda — but that acknowledgment has not produced any concrete instrument.
The AI Council itself has noted, 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.
What the Profile Shows
Tajikistan in 2023 held an unusual position: first in the region with a national AI strategy, and first low-income country globally to adopt one — while also being the country where the gap between that strategy and a responsible AI framework was widest. The technology is deploying, the GDP targets are set, and the governance infrastructure is staged for later. The GIRAI baseline captures that sequence. Whether the accountability layer arrives before or after the systems it is meant to govern is what subsequent assessments will be able to measure.
Based on the Tajikistan country context and research findings submitted to the Global Index on Responsible AI (GIRAI) 1st Edition, 2023. Data source: global-index.ai. Strategy reference: Digital Watch Observatory. Regional hub: IDFI (Georgia). This is an observational read by KG Labs as part of its Central Asia AI governance coverage. Publication consent: pending author confirmation.
