The Philippines Just Set Rules for AI in Schools. Here's What Every SEA Educator Should Know.
It is a Tuesday afternoon in Cebu, and a Grade 9 science teacher is doing something she never imagined doing two years ago: she is asking an AI to help her draft a quiz on photosynthesis. Not because her school told her to. Not because she has had any training. Because her students are already using AI to complete their homework, and she figures she had better understand it before they know more than she does.
She is not alone. Across Southeast Asia, teachers at every level are improvising their relationship with artificial intelligence: some embracing it, some banning it, most navigating the grey space between. And for years, the policy vacuum has made that grey space enormous.
In June 2026, the Philippines did something that no other K-12 education system in Southeast Asia had done before it: it published a formal, named, tool-specific set of AI guidelines for schools. DepEd Order 003, Series of 2026 (the Department of Education's first official AI framework for K-12) does not just gesture at "responsible use." It names specific tools. It sets ethical standards. It puts human oversight at the centre. And it offers a template that every educator in the region, whether or not they teach in the Philippines, should read carefully.
What DepEd Order 003 Actually Says
The order runs to several pages of policy language, but its core commitments are straightforward enough to sketch here.
First, it permits the use of named AI tools (ChatGPT, Google Gemini, and Khanmigo among them) in teaching and learning contexts. This is a significant step. Rather than issuing a blanket ban (as some institutions have done) or a blanket permission (which creates its own chaos), DepEd chose to name specific platforms that meet its standards and set conditions for their use.
Second, it establishes that AI must be used under ethical, human-centred standards. This means AI tools are positioned as aids to educators, not substitutes for them. Teachers remain accountable for what happens in their classrooms, AI-assisted or otherwise. The framing is deliberate: the human is in the loop, always.
Third, it requires that AI use be transparent. Students should know when AI has been involved in their learning or assessment. Educators should be able to explain why a given tool is being used and what purpose it serves.
Fourth, and this is the part that deserves more attention, it acknowledges that AI in education is not a technology problem. It is a pedagogy problem. The question is not which tool to use. It is what learning outcome you are trying to achieve, and whether AI genuinely helps you get there.
Why This Matters Beyond the Philippines
The Philippines is educating roughly 28 million K-12 students, making it one of the largest school systems in ASEAN. A policy shift at this scale sends signals across the region.
But the significance is less about size and more about sequence. Singapore has focused its AI literacy efforts heavily at the tertiary level. Its AI in Education framework, while thoughtful, has largely addressed universities and polytechnics before trickling downward. Malaysia announced in late 2025 that it would deploy Google Gemini across public universities, again a tertiary-first approach. Indonesia has piloted AI tools in vocational education. Vietnam is exploring AI for teacher training.
The Philippines is the first in the region to draw a formal line at K-12 and say: this age group, these students, these classrooms: here are the rules.
That matters because the stakes at K-12 are different. Younger students are more susceptible to misinformation. The dependency risks are higher. The equity gaps are wider. And the teachers, many of whom are managing classes of 50 or more students with minimal support, are more likely to encounter AI without adequate preparation than their university counterparts.
DepEd Order 003 is not a perfect document. It will need updating as the technology moves. But it is the most concrete attempt in SEA to answer the question that every teacher in the region is actually asking: can I use this, and if so, how?
What "Human-Centred AI" Means in a Real Classroom
The phrase "human-centred AI" appears in many policy documents and almost none of them define it in terms a classroom teacher can act on. DepEd Order 003 brings it closer to earth, but it is still worth unpacking what it actually looks like on a Tuesday afternoon.
Human-centred AI, in a teaching context, means the educator retains authorship. The curriculum is yours. The learning outcomes are yours. The assessment criteria are yours. The AI does not decide what your students should know or how they should be evaluated. It helps you deliver the learning experience you have already designed.
It means the AI is bounded. It is not a general-purpose search engine that a student can use to find anything on the open internet. It works within the scope of what you have defined: the course content, the syllabus, the types of questions that are appropriate.
And it means feedback is pedagogically grounded, not just statistically generated. When an AI tells a student their answer is incorrect, the correction should be rooted in the learning objective, not just in pattern-matching against a training set.
This is harder to build than it sounds. Most general-purpose AI tools are not designed with these constraints in mind. They are designed to answer any question, helpfully and fluently, regardless of whether the answer serves the student's actual learning.
Platforms like Noodle Factory's Walter are built around a different premise: educators define the content, the outcomes, and the guardrails. Walter does not roam the open internet. It teaches within the boundaries the educator sets. The AI is, in a meaningful sense, the teacher's instrument, not a free agent. That alignment with the spirit of DepEd Order 003 is not coincidental. It reflects a conviction about where AI belongs in education: as a capable assistant operating under human direction, not as an autonomous instructor.
The Honest Caveats
A policy announcement is not an implementation. Between DepEd Order 003 and a Grade 9 science teacher in a rural barangay confidently using an AI tool in her lesson lies a considerable distance.
Infrastructure gaps are real. The Philippines' National ICT Household Survey has consistently shown uneven internet penetration, particularly in Mindanao and the Visayas outside major cities. An AI policy that assumes reliable connectivity will leave the most under-resourced schools behind.
Teacher preparation takes time. Naming ChatGPT in a policy document does not tell a teacher how to use it, when not to use it, or how to respond when a student uses it to cheat. Professional development at scale, for a teaching workforce of roughly 900,000, is a different order of challenge than writing the policy itself.
Language is complicated. The Philippines has over 180 languages and regional dialects. Most AI tools perform significantly better in English and Tagalog than in Cebuano, Ilocano, or Hiligaynon. A framework that works for Manila may not work the same way for Davao.
None of this is reason to dismiss the policy. It is reason to hold it to account: to insist that implementation funding, teacher training, and infrastructure investment follow the paper.
What Educators Elsewhere Should Take From This
If you are a teacher or school leader in Malaysia, Indonesia, Vietnam, Thailand, or Singapore, DepEd Order 003 is relevant to you even if it has no legal force in your context. It is the most concrete public articulation of what an ethical AI framework for K-12 schools looks like in this region.
Three things are worth borrowing, regardless of your jurisdiction.
The naming principle. Rather than issuing vague guidance about "approved tools," DepEd named specific platforms and evaluated them against criteria. Your school or institution can do this too. It is more work upfront, but it gives teachers clarity and accountability.
The transparency norm. Making AI use visible (to students, to parents, to administrators) reduces the conditions under which AI becomes a tool for academic dishonesty and increases the conditions under which it becomes a genuine learning aid. Transparency is a practice, not just a principle.
The pedagogy-first framing. The best AI policies in education start with learning outcomes, not technology features. What do you want students to be able to do? How does AI help them get there? What does the educator need to stay in control of? These are pedagogical questions, and they should precede every tool selection decision.
The teacher in Cebu drafting her photosynthesis quiz deserves better than a policy vacuum. She deserves a framework that tells her what is permitted, why it matters, and how to use it well. The Philippines just gave its teachers that framework. The rest of the region is watching.
Further Reading
Sources
Department of Education, Philippines. DepEd Order 003, Series of 2026: Guidelines on the Ethical and Responsible Use of Artificial Intelligence in Basic Education. June 2026.
Philippine Statistics Authority. Functional Literacy, Education and Mass Media Survey (FLEMMS). 2022.
Department of Education, Philippines. Basic Education Statistics. 2025.
Ministry of Education, Singapore. AI in Education Framework. 2024.
Ministry of Education, Malaysia. Gemini for Public Universities Deployment Announcement. 2025.
UNESCO. AI Competency Framework for Teachers. 2024.
SEAMEO. Status of ICT Integration in Education in Southeast Asian Countries. 2023.
Google. Khanmigo Product Documentation. 2025.
