[The End of Coding Education] “Mandatory AI From Age Three”: UAE’s Education Drive Risks Falling Into the ‘Early Coding’ Trap

Securing Post-Oil Growth Engines: The UAE’s Mandatory AI Education
Coding and Data Analytics Prioritized Before Advanced Mathematics
Risk of Shallow Conceptual Understanding and Widening Learning Gaps

The United Arab Emirates (UAE) is accelerating its nationwide talent development drive by making artificial intelligence (AI) education mandatory from kindergarten through high school. The initiative forms part of a strategy to secure new growth engines for the post-oil era, but concerns are mounting that prioritizing coding and data analytics without a mathematical foundation could widen learning gaps among students with weak basic academic skills. With generative AI now capable of writing code, critics warn that early coding education could encourage students to imitate examples without developing the ability to use technology effectively. South Korea’s experience—where private education spending has reached roughly $20 billion while problem-solving proficiency remains below the Organization for Economic Cooperation and Development (OECD) average—offers a stark warning about the risks of the UAE’s accelerated education drive.

Mandatory AI Classes from Kindergarten Through High School

According to the UAE Ministry of Education on August 10, the country is pursuing an ambitious initiative to establish itself as a technological superpower by building a comprehensive AI education system spanning early childhood education through postgraduate study. Since May last year, the UAE government has required all public schools to provide AI classes from kindergarten through high school. Children as young as three and four have consequently begun encountering basic AI concepts. The Ministry of Education aims to equip students with the essential AI knowledge and skills needed to succeed in a rapidly evolving digital world.

The National AI Curriculum Framework developed by the Ministry of Education divides the curriculum into three learning domains, seven strands, and three educational stages. The introductory stage covers conceptual awareness and ethical responsibility, while the advanced stage requires students to design, evaluate, validate, and apply AI systems in real-world settings. Drawing on student AI competency frameworks established by the United Nations Educational, Scientific and Cultural Organization (UNESCO) and the OECD, the curriculum incorporates critical thinking, problem-solving, and ethical judgment into its educational objectives.

An AI Talent Pipeline Combining Coding and Data Analytics

In kindergarten, for example, children encounter foundational AI concepts through stories and play. Students from the lower elementary grades through middle school study coding, data analytics, practical AI applications such as voice assistants, and related ethical issues. At the high school level, students tackle advanced subjects including machine learning (ML) and artificial neural networks while completing project-based assignments such as developing chatbots and analyzing datasets. The curriculum is designed to build practical capabilities that can lead to university study or careers in technology.

Ten-year-old students at UAE public schools are already learning not only the fundamentals of programming but also methods for preventing deepfakes and hacking and identifying AI-generated images. At institutions such as the GEMS School of Research and Innovation, which opened last August, children aged five and six learn coding and robot-control algorithms before progressing toward adult-level data science and robotics skills.

The UAE has spent years laying the groundwork for this initiative, appointing the world’s first minister of state for AI in 2017 and investing heavily in AI research and startups. The decision to mandate AI education extends this national innovation strategy and reflects the government’s determination to integrate future technologies throughout the education system. Under its National Strategy for Artificial Intelligence 2031, the UAE is deploying AI across public administration, healthcare, transportation, energy, and finance. This year, it also announced a goal of converting 50% of federal government functions, services, and operational processes to agentic AI systems over the next two years. An AI data center complex being developed by the United States and the UAE in Abu Dhabi is expected to reach a final power capacity of 5 gigawatts (GW). Advanced semiconductors, data centers, administrative automation, and education curricula are being integrated into a unified national industrial policy.

The High School Mathematics Barrier Confronting Elementary Coding

Experts nevertheless emphasize that AI education and coding education should not be treated as interchangeable concepts. Block-based coding and conditional statements can be learned before high school mathematics, but designing AI models and understanding data computation and training processes require a substantially stronger foundation. Working with data requires knowledge of matrices, while designing inputs and outputs requires an understanding of functions. Derivatives are used to reduce errors in AI models, and exponential and logarithmic functions, probability, and statistics are also essential. Experts say it is virtually impossible for elementary school students who have yet to learn even factorization to write and modify code based on these principles independently.

This is why Stanford University’s flagship machine-learning course, CS229, lists probability theory, multivariable calculus, and linear algebra as prerequisites. Google’s Machine Learning Crash Course similarly identifies algebra, linear algebra, and statistics as essential foundations and recommends calculus for advanced study. Modern AI relies on this academic foundation: data are represented through matrix operations, inputs and outputs are connected through functions, and errors are reduced through probability distributions and differentiation. When AI coding is taught first to elementary school students who have not learned this mathematics, only a small minority are likely to grasp the underlying principles, while the rest may merely copy and enter sample code.

The foundational academic proficiency of UAE students also remains fragile relative to the demands of AI education across all grade levels. In the OECD’s Programme for International Student Assessment (PISA) 2022, 51% of UAE students attained at least Level 2 proficiency in mathematics, compared with an OECD average of 69%. Reading proficiency stood at 52%, below the 74% average, while science proficiency reached only 55%, well short of the 76% average. Nearly half of 15-year-old students had particular difficulty understanding basic mathematical representations and identifying the central ideas in medium-length passages.

South Korea Training ‘Coders’ in the ‘Architect Era’

South Korea’s experience illustrates the costs that can accompany a race to accelerate AI education. Generative AI can already produce complex code within seconds and identify errors. The role demanded in software development is shifting from the “coder,” who enters code manually, to the “architect,” who uses AI to design entire systems.

South Korea’s private education market, however, continues to move in the opposite direction. According to the Seoul Gangnam-Seocho District Office of Education, the number of coding and robotics academies in Seoul’s Gangnam, Seocho, and Nowon districts and Bundang in Seongnam, Gyeonggi Province, increased 39% from 90 in 2022 to 125 last year. Some institutions charge as much as $423 per month for two hours of classes per week, excluding educational materials, while enrollment now extends to kindergarten-age children.

Anxiety generated by education policy has also fueled the expansion of the private academy market. The government repeatedly announced plans to expand AI-focused schools and curricula but failed to specify the content and difficulty level for each grade, deepening the information vacuum confronting parents. Private academies moved quickly into that gap, promoting advance Python classes and robotics courses with claims that students must first learn coding to use AI effectively. They also advertised that students need prior coding experience to enroll in introductory AI and data science courses under the high school credit system and that related projects could strengthen their academic records. Parents find it difficult to measure AI literacy, while coding produces immediately visible results on a screen. This reflects a familiar private education sales model that converts uncertainty about the future into tangible certificates and competition records.

The educational lifespan of memorizing coding syntax is rapidly contracting. Shin Jong-ho, a professor of education at Seoul National University, identified computational thinking—the ability to break problems into smaller units, identify patterns, and design solutions—as a core competency for the AI era. He also stressed the need to cultivate the ability to define problems independently, think critically when formulating questions for AI, assess errors in outputs, and collaborate effectively. Education centered on memorizing Python and C syntax is increasingly disconnected from a reality in which AI writes the code itself.

Table 1. The Gap Between Educational Inputs and Problem-Solving Capabilities in South Korea

South Korean Education Trapped in a High-Cost, Low-Efficiency Quagmire

The early-coding frenzy is rooted in the answer-driven competition that has long dominated South Korean education. Students have grown accustomed to training that rewards finding predetermined answers quickly, with less emphasis placed on independently formulating questions and designing solutions. Whenever a new technology emerges, South Korea first examines whether it will influence university admissions before assessing its educational value. Once other students begin learning it, fear of falling behind quickly takes hold. The convergence of AI education toward advance Python study, certificate acquisition, and competition preparation is another product of this competitive system.

The same limitations are evident in the data. In the PISA 2022 creative thinking assessment, 46% of South Korean students reached the highest proficiency levels, far exceeding the OECD average of 27%. In the OECD Survey of Adult Skills, however, 37% of South Korean adults remained at Level 1 or below in adaptive problem-solving, while only 1% reached the highest level. Students demonstrate strong capabilities in completing assignments defined by schools, yet the ability to frame complex real-world problems, select relevant information, and devise solutions fails to carry through into adulthood.

Massive education spending has also failed to close this gap. Private education spending for South Korean elementary, middle, and high school students reached a record $20.61 billion in 2024, with the participation rate rising to 80%. The total declined to $19.41 billion in 2025, partly because of a shrinking student population, but average monthly spending per participating student rose to a record $426. The OECD also noted in its 2026 Economic Survey of Korea that adult information-processing proficiency remained below the member-country average despite South Korea’s substantial investment in formal and private education. A system built around intensive and prolonged study has failed to deliver problem-solving capabilities commensurate with its high educational costs.