Table of Contents
Teach AI in schools used to be a debate about whether to do it at all. In 2026 the question flipped: the race is about who does it first, how early, and whether the system survives. One country now teaches AI as an official subject from kindergarten to Grade 12, one treats it as national industrial policy, one tried to automate the classroom and got stopped by its own parliament, and one small European nation is quietly running the most sophisticated experiment of all. This is the country-by-country scoreboard for the race to teach AI in schools — and the verdict on where the Philippines stands today.
The scoreboard on who gets to teach AI in schools matters because the debate phase is over. New York restricted school AI use while China made instruction compulsory; that confrontation is covered in our AI-in-schools policy piece. What follows here is the different question parents, teachers and policymakers are now asking: not “should we,” but “who is actually building the machine that teaches AI” — and what it costs the countries that get it wrong.
Key Takeaway
- 🏆 The UAE holds the breadth record: the first country to make AI an official school subject from kindergarten to Grade 12, starting the 2025-26 academic year, inside existing class time.
- 🇨🇳 China leads on speed and depth: AI instruction made compulsory from elementary school in leading cities, with at least 10 hours a year in Hangzhou’s primary grades — and “verify the answer” written into the curriculum itself.
- 📉 South Korea is the cautionary tale: AI textbooks went national in March 2025, were stripped of official textbook status by parliament in August, and adoption collapsed from 37% to 19% of schools.
- 🧪 Estonia runs the most sophisticated integration: roughly 20,000 students and 3,000 teachers get structured AI access through Socrates AI, built with OpenAI — guidance-first, not answer-first.
- 🇵🇭 The Philippines is leading the teacher-first lane in Southeast Asia — but has no mandatory AI subject: AGAP.AI reaches up to 1.5 million Filipinos, DepEd Order No. 003 s. 2026 is the first national AI policy, and the Google AI Educator Series launched October 13, 2026.
The scoreboard: who is winning the race to teach AI in schools
The comparison only works if it counts the same things for every country: how early instruction starts, whether it is mandatory or optional, whether AI is a subject or a tool, and whether the rollout survived contact with parents and teachers. Here is the standing board.
| Country / city | What is taught | Grade span | Mandate status |
|---|---|---|---|
| UAE | Official AI subject: seven learning areas from foundational concepts to ethics and project design | Kindergarten (age 4) to Grade 12 | Mandatory, all public schools, since 2025-26 |
| China (leading cities) | AI literacy plus critical verification of AI answers | Elementary (in Hangzhou: at least 10 hours a year) through secondary | Compulsory in leading cities; national strategy |
| South Korea | AI digital textbooks (math, English, informatics) | Upper primary to secondary | Rolled back: demoted to supplementary material, optional per school |
| Estonia | Socrates AI — guided AI use across subjects | Focus on 16-year-old high-school students and their teachers | National program, ~20,000 students, ~3,000 teachers |
| Singapore | Personal digital device for every middle-school student; teacher-led AI learning | Middle school onward | Systematic expansion |
| Boston, USA | AI curriculum in public high schools | Secondary | First in the US, classes started September 2026 |
| Philippines | Teacher-first: AGAP.AI training, Google AI Educator Series, MIT Day of AI curriculum work, e-TURO classroom pilots | Basic education (teacher + integration layer first) | Policy exists (DepEd Order No. 003 s. 2026); no mandatory student subject yet |
Read the mandate column carefully — it is where the race to teach AI in schools actually separates. Most national systems can announce an AI initiative; very few changed the timetable of a child’s school day.
The UAE: the first full-ladder mandate
The United Arab Emirates made AI an official subject across the entire public-school system — kindergarten at age four through Grade 12 — starting the 2025-26 academic year. The UAE’s national AI literacy framework defines seven learning areas: foundational concepts, data and algorithms, software use, ethical awareness, real-world applications, innovation and project design, and policies and community engagement. Kindergarten children start with visual, story-based activities; by the later grades, the work moves to applied projects. Crucially, the design avoids the classic rollout failure: AI lands inside the existing Computing, Creative Design and Innovation timetable, with no additional teaching hours — the ministry paired it with delivery partners including Presight (a G42 company), AI71, Mohamed bin Zayed University of Artificial Intelligence and the Emirates College for Advanced Education.
Why the UAE matters beyond its size: it built the full ladder every other country in the race to teach AI in schools starts from the middle of. Every other system in this piece starts somewhere in the middle — a high-school course, a device program, a pilot. The UAE made the declaration that AI literacy begins at four, which forces the entire supporting machine — teacher training, materials, assessment — to stand up at every grade at once. That is the broadest mandate on the board, benchmarked by its own ministry against UNESCO’s AI competency framework for students and the OECD/European Commission framework.
China: AI as industrial policy, taught from elementary
China is running the same race at industrial-policy speed. In leading cities — Hangzhou and Shanghai are the documented examples — AI instruction is compulsory from elementary school, with Hangzhou’s primary students required to receive at least ten hours of AI instruction a year. The distinctive move is not the hours; it is the content. Reports on the Chinese approach note that the ability to question and verify AI answers has been written into the curriculum as a required skill — critical use, not just tool use.
It pairs with the state machinery that lets China teach AI in schools at industrial speed: a national AI education push, dedicated AI textbooks and platforms, and a talent pipeline logic (the AI race we tracked at the country level in the China versus US piece extends into primary classrooms). The honest caveat for the scoreboard: city-level mandates lead the way, with national coverage phased — which is why China appears as the depth-and-speed leader rather than a uniform K-12 mandate like the UAE.
South Korea: the cautionary tale every ministry should read
South Korea tried the boldest version of “AI in the classroom”: government-built AI digital textbooks in mathematics, English and informatics, introduced starting March 2025. Within one school year the project hit the wall. Parents and teachers pushed back over inaccuracies, data-privacy risks and workload; lawmakers formally revoked the textbooks’ official status in August 2025, demoting them from textbooks to “supplementary material” — adoption left to each school’s discretion. The numbers tell the collapse plainly: adoption fell from 37% of schools in the first semester to 19% by September, with about 2,095 schools still signed on, roughly half the level earlier in the year.
The lesson for any ministry that wants to teach AI in schools is not “AI textbooks failed” — it is that a system imposed at national speed without parent trust, privacy guarantees and teacher capacity will be stopped, even in one of the world’s most tech-forward education systems. Every ministry that wants to teach AI in schools now inherits Korea’s checklist: consent, accuracy, workload, exit ramps. A classroom decision about a seven-year-old’s schooling turned political in months — the AI-textbook champion’s own administration was already gone when the rollback came.
Estonia and Singapore: the integration lane
Estonia — PISA’s third-ranked country in the world and Europe’s strongest — is running the most sophisticated experiment on the board. Through AI Leap and the Socrates AI program, built in partnership with OpenAI and other AI companies, roughly 20,000 students and 3,000 teachers received free, structured AI access: an adapted ChatGPT that guides students rather than handing them answers. Education Minister Kristina Kallas stated the philosophy directly — “I don’t think banning AI will prevent young people from using it” — the national bet is that teaching disciplined use beats policing prohibited use — the quiet way to teach AI in schools without a mandate.
Singapore takes a device-per-student path: every middle-school student has a personal digital device, but the learning is designed around interaction with teachers rather than screen-to-screen. The model keeps the human at the center while making AI a daily, ordinary tool — the opposite failure mode from Korea’s stand-alone automated textbooks. Both countries teach the same meta-skill: how to use AI critically and ethically, which is exactly the content China also decided to mandate.
The United States and the late movers
The US has no federal mandate — the race to teach AI in schools runs through cities and states. Boston became the first US city to introduce an AI curriculum in its public high schools, with classes starting September 2026. Analysts read it as the start of AI instruction in American public education, pushed by the same talent race that drives China’s elementary mandates and the UAE’s universal ladder. The gap between these moves and the rest of the US states is now a policy conversation of its own — one more version of the access divide, where AI instruction can arrive in private systems years before public ones.
The three strategies in the race to teach AI in schools
Strip the country names away and the race to teach AI in schools has three strategy families:
- Mandate: make AI a subject with hours attached (UAE K-12; China’s leading cities). Fast coverage, high political cost, needs a standing teacher-and-materials machine.
- Build: put AI inside the textbooks and platforms themselves (South Korea’s original plan, parts of China’s edtech stack). Deepest integration, highest failure risk — Korea is the proof.
- Augment: give the teaching force and students disciplined access and let integration grow subject by subject (Estonia’s Socrates AI, Singapore’s device model, Boston’s course path, the Philippines’ teacher-first programs). Slowest-looking, most durable.
The 2026 scoreboard verdict: the UAE leads the mandate lane, China leads the industrial-speed lane, Estonia leads the integration lane — and South Korea owns the most-watched case study of what destroys a rollout. No country leads all three lanes, for the same structural reason the AI infrastructure race has no single winner: classrooms are systems, and systems can only move as fast as their teachers.
WorldNgayon Insight: what the race is really deciding
What changed: the question flipped from whether AI belongs in schools to who reaches which grade first — 2026 produced the first full-ladder mandate and the first public rollback. Why now: the 2023-25 classroom AI explosion outpaced policy; 2026 is when legislatures and ministries forced the issue. Why it lasts beyond the news cycle: curriculum decisions compound — a generation that learns to verify AI answers in Grade 4 does not grow up into the same workforce as one that learned at 19.
Who is affected: every parent weighing screen-time rules, every teacher absorbing new training, every student whose curriculum now includes a machine that answers back. What to do next: read your country’s lane — mandate, build or augment — and judge it against Korea’s checklist before believing any announcement. Risks, opportunities, costs, skills: the risk is the Korea path (speed without consent); the opportunity is Estonia’s (integration with consent); the skill everyone is converging on is critical verification of AI output. Connection to work and money: the countries that teach AI in schools early are building the hiring pool for the AI-skilled roles our skills-demand analysis tracks — the classroom wage gap of the 2030s is being drawn now.
One verdict sentence, earned from the board: the winner will not be the country that announced first — it will be the one whose teachers are still standing in the classroom when the textbooks stop being experimental.
Where the Philippines stands today
The Philippines chose its lane in the race to teach AI in schools deliberately: the augment lane, and with surprising speed this year.
The teacher-first program is real and dated. DepEd’s Project AGAP.AI (Accelerating Governance and Adaptive Pedagogy through Artificial Intelligence) was launched with President Ferdinand Marcos Jr. personally on stage — a national AI training push for basic education reaching up to 1.5 million Filipinos, including learners, teachers and parents, run with the ASEAN Foundation and Google.org. The courses cover responsible use, misinformation detection, privacy and security, and AI literacy.
The national policy exists. DepEd Order No. 003, s. 2026 is the first comprehensive AI policy for basic education: teachers may use AI for lesson planning, quiz generation, content development and feedback, while AI cannot be the sole basis for grading, learner evaluation or major academic decisions — human judgment stays at the center. The order itself acknowledges that classroom AI adoption had already outpaced the system’s rules.
The curriculum work is contracted. DepEd is working with MIT (Responsible AI for Social Empowerment and Education) and the Day of AI program to integrate AI concepts into the basic-education curriculum — this is the track that, if it completes, moves the Philippines from the augment lane toward a real subject ladder.
The classroom pilots and the private sector are moving. DOST-CAR and DepEd pilot the e-TURO AI learning system in Tabuk public schools; Google’s AI Educator Series launched October 13, 2026 with DepEd, CHED and TESDA — free, self-paced modules with NEAP-accredited CPD credit, extending Google’s effort to train one million DepEd educators. The Google launch study found 86% of Filipino educators inclined to use AI — and flagged training as the gap, which is precisely the lane the country is running.
The honest verdict. In Southeast Asia’s part of the race to teach AI in schools, — where Singapore runs its device program and no ASEAN neighbor yet has a UAE-style mandate — the Philippines is the region’s most aggressive teacher-first bet: more than a million educators being trained before any subject is made compulsory. The gap on the scoreboard is equally clear: no mandatory AI subject, no grade-span ladder, national readiness still uneven (power, devices, connectivity in far-flung schools — the constraint map from our Southeast Asia AI infrastructure race piece applies to classrooms too). The Philippines is not winning the mandate race. It is trying to win the race that Korea lost consent for and Estonia won trust through — the race for a teaching system that survives. The 2027 signal to watch: whether the MIT/Day of AI curriculum integration graduates from contract to actual DepEd learning materials, and whether AGAP.AI’s 1.5-million figure converts into completed training records.
What changes the scoreboard next
- China’s city mandates going national — if Beijing extends the elementary mandate nationwide, the depth lane goes from phased to universal.
- South Korea’s supplementary era — whether new rules rebuild parent trust and adoption recovers, or the program becomes the default warning slide in every ministry deck.
- The Philippines’ contract-to-curriculum moment — MIT/Day of AI integration becoming official DepEd materials would be the country’s first ladder step.
- Boston’s first-semester data — the first US public city program becomes either the domestic template or another rollback case.
Who was first to make AI a mandatory school subject?
The UAE was the first country to make AI an official subject across the full public-school ladder — kindergarten (age 4) to Grade 12 — starting with the 2025-26 academic year, delivered inside existing subject time with no added hours.
At what age does China teach AI in schools?
In leading cities, AI instruction is compulsory from elementary school; Hangzhou requires at least 10 hours per year in primary grades, with Shanghai running parallel mandates. Verification of AI answers is itself part of the required curriculum.
Why did South Korea roll back its AI textbooks?
Parents and teachers pushed back against inaccuracies, data-privacy risks and increased workload after the March 2025 national rollout; in August 2025 lawmakers revoked the books’ official textbook status, demoting them to optional supplementary material. Adoption fell from 37% to 19% of schools within two semesters.
Is the Philippines teaching AI in schools?
Not as a mandatory subject yet. The Philippines runs the augment lane: AGAP.AI teacher-first training for up to 1.5 million Filipinos, the first national AI policy for basic education (DepEd Order No. 003, s. 2026), MIT and Day of AI curriculum-integration work, e-TURO classroom pilots, and the Google AI Educator Series launched October 13, 2026.
Which country has the best model for using AI responsibly in class?
Estonia’s Socrates AI — free, structured access for ~20,000 students and ~3,000 teachers through an adapted, guidance-first ChatGPT built with OpenAI — currently reads as the most mature integration model, in a PISA top-three system.
What should parents ask their school this year?
Whether the school has adopted a written AI policy (like the Philippines’ DepEd Order No. 003), whether AI use teaches verification of AI answers or just tool use, and what the exit ramps are when a tool fails — Korea’s rollout showed those three checks decide whether a program survives.
Does teaching AI in schools mean replacing teachers?
No system on the 2026 scoreboard teaches AI by removing the teacher — Singapore’s design rule is explicit interaction with teachers over devices; the Philippines’ policy keeps grading and evaluation with humans; Estonia’s tool guides rather than answers. The race is about training teachers, not automating them.
What is the fastest signal that a country is pulling ahead?
The mandate column, not the announcements: whether AI shows up in a child’s official timetable — subjects, hours, grade span — rather than in press releases. That is the single number this scoreboard tracks on the race to teach AI in schools.
Sources: UAE Ministry of Education — National AI Literacy Curriculum Framework (K-12); WAM — UAE MoE announcement, May 4, 2025; Rest of World — South Korea’s AI textbook rollback; Seoul Economic Daily — Boston and the global race; El País — Estonia, New York, and the global classroom debate; Philippine Information Agency — AGAP.AI launch (pia.gov.ph); Inquirer — Google AI Educator Series, October 13, 2026; PIA — e-TURO pilot in Tabuk (pia.gov.ph). Figures are as reported by the named sources; policy claims trace to the issuing ministry or legislature’s own record where linked.
