AI in Uzbekistan: A Question of Manageability, Not Implementation
Analytical Note · Series: Governance Briefs · 1 of 9 Operationalises: Essay 7 “The Correction Window”, Essay 8 “The Agency Transfer”, Essay 11 “The Institutional Gap”
Русская версия: okhodjaev.com/governance-briefs/the-manageability-question-ru/
Over recent years, Uzbekistan has established a regulatory foundation for artificial intelligence development that few countries in the region — and relatively few in the wider developing world — can match: the AI Technology Development Strategy to 2030 (Presidential Decree PP-358, 14.10.2024), legislative amendments introducing the first legal definition of AI and restrictions on fully autonomous legally significant decisions (Law ZRU-1115, 21.01.2026), Ethical Rules for AI Application (Order of the Ministry of Digital Technologies No. 3787, 14.03.2026), a target of 100 priority projects, and financing of AI-related projects exceeding $150 million through the Reconstruction and Development Fund of Uzbekistan. This represents substantial institutional work carried out within a compressed timeframe.
The next challenge is structured quite differently.
Regulatory documents define what is permissible. They do not yet create instruments for verifying the actual behaviour of AI systems in operation. One of the central limitations of the new legislation is the prohibition on sole reliance on AI-generated conclusions when taking legally significant decisions that affect the rights and freedoms of citizens (Article 7¹, ZRU-1115). The norm is legally sound in its intent.
The norm exists — the instrument for its enforcement does not.
In the current architecture, there is no definition of what constitutes “exclusive” reliance, no mechanism for verifying compliance, and no established procedure for responding to identified violations.
This is not an oversight by the drafters. It is a structural characteristic of first-generation AI regulation: the normative framework is created faster than the mechanisms for its operationalisation.
International practice in banking supervision, pharmaceuticals, and nuclear verification reveals a consistent pattern. Where the function of developing a technology and the function of independently assessing its behaviour in operation are concentrated within the same institutional structure, a divergence gradually emerges between the logic of technology promotion and the logic of independent oversight. States that separated these functions at an early stage retained real — not merely legal — control. Those that delayed the separation found that key system parameters had already taken shape outside the original design intent.
In the field of AI, the window for corrective action is narrowing significantly faster than in any other sector. Systems accumulate operational history, and each additional period of operation without a verification protocol increases the cost of subsequent audit — while simultaneously reducing the practical significance of its findings. As dependency deepens, management expertise gradually shifts from understanding how decisions are formed to managing only their outcomes. Uzbekistan’s banking sector, actively deploying AI-based credit scoring, stands at the beginning of this curve. The point at which rollback costs begin to grow non-linearly has not yet been reached — but it is not as far off as is generally assumed.
For Uzbekistan, this defines a concrete task for the next stage of regulatory maturity. What is needed is not simply more documents, but documents of a different kind: a mechanism that functionally separates the development, financing, and promotion of AI projects from their independent assessment; a verification standard that enables supervisory bodies to examine the real-world behaviour of a system rather than its declared characteristics; and a response protocol for AI systems embedded in critical infrastructure — one that exists before the first serious incident, not after.
As AI integration deepens, the cost of building such an architecture will only grow. The experience of other sectors is unambiguous: effective oversight is built before the onset of deep operational dependency — not after.
📄 Download PDF (English) · Скачать PDF (русский)
Sources
[1] Presidential Decree of the Republic of Uzbekistan No. PP-358, 14.10.2024, “On Approving the AI Technology Development Strategy to 2030.” lex.uz
[2] Law of the Republic of Uzbekistan No. ZRU-1115, 21.01.2026, “On Amendments to Certain Legislative Acts in Connection with the Regulation of Relations Arising from the Use of Artificial Intelligence.” lex.uz
[3] Law of the Republic of Uzbekistan No. 560-II, 11.12.2003, “On Informatisation”, as amended, including Article 7¹ introduced by ZRU-1115. lex.uz
[4] Order of the Ministry of Digital Technologies of the Republic of Uzbekistan No. 3787, 14.03.2026, “On Approving Ethical Rules for the Development, Implementation and Use of AI-Based Solutions.”
[5] Presidential Decree of the Republic of Uzbekistan No. UP-189, 22.10.2025, “On Additional Measures for the Further Development of Artificial Intelligence Technologies.” lex.uz
[6] Presidential Resolution of the Republic of Uzbekistan No. PP-320, 30.10.2025, “On Additional Measures for Supporting Projects Based on Artificial Intelligence Technologies.” lex.uz
[7] Presidential Resolution of the Republic of Uzbekistan No. PP-4996, 17.02.2021, “On Measures for Creating Conditions for the Accelerated Implementation of Artificial Intelligence Technologies.” lex.uz
[8] Basel Committee on Banking Supervision. Principles for Sound Management of Operational Risk. Bank for International Settlements. bis.org
Full brief and updated sources: okhodjaev.com/governance-briefs/the-manageability-question/
Oybek Khodjaev — over 35 years of experience in banking, finance, public administration and business in Uzbekistan and the CIS. Author of the essay series “Beyond Control: The Theory of the Limits of AI Governance”. okhodjaev.com
The author advises public institutions and financial organisations on AI governance, verification frameworks, and institutional readiness.