Summary: On or around February 12, 2026, the UAE Ministry of Education issued a manual titled “2026 Safe and Responsible Use of Artificial Intelligence in Classrooms,” setting firm boundaries for generative AI in schools: no student under 13, or who has not yet reached Year 7, may use generative AI tools at all. If you build education products, treat the list as a requirements document in disguise: age gating, content classification, data protection, and platform allowlists each translate into a feature you can actually ship.
On or around February 12, 2026, the UAE Ministry of Education issued a manual titled “2026 Safe and Responsible Use of Artificial Intelligence in Classrooms,” setting firm boundaries for generative AI in schools: no student under 13, or who has not yet reached Year 7, may use generative AI tools at all. Both Gulf News (its report updated February 14) and Times of India covered the document, the latter building its headline around the “25 new AI guidelines.”
What separates it from yet another set of aspirational principles is that the rules operate at classroom level and can actually be enforced — and the timing is deliberate: from August 2026, every government school and every private school on the MoE curriculum must teach AI education from kindergarten through Grade 12, across seven core areas. Running a ban and a mandate in parallel makes this the most concrete specimen of K-12 AI governance to appear so far in 2026 (see our 2026 opening outlook). Its arrival, while regulators elsewhere are still arguing over definitions, is precisely why it will get studied.
The Core Provisions
Age is the boundary the whole manual turns on. The ministry justifies it as protection for younger learners — that stage of education is meant to stay anchored in “interaction, creativity and independent skill development.” Whatever the stated purpose, no student under 13 or in a grade below Year 7 may use generative tools.
The remainder of the document fans out into roughly 25 prohibitions. The toughest targets assessment: AI is off-limits during formal exams and official assessments. Any assignment, piece of research, or report produced with AI must be disclosed and signed off by a teacher; passing it off as personal work without either is a violation. Copying or rephrasing AI output with no genuine understanding behind it is banned on its own.
Reading the Red Lines
Sort them and four groups emerge. The first covers content: nothing involving violence, hate speech, discrimination, misinformation, or psychological harm may be generated or shared, and nothing may conflict with “religious, national or cultural values”; material encouraging cyberbullying, self-harm, or illegal behavior is excluded outright. The second covers personal data: names, photos, audio, video, and contact details of students, teachers, or parents may not be fed into AI systems. The third covers safety: deepfakes, impersonation, harvesting sensitive information about others, reaching into confidential school materials, and recording or transcribing classes without the consent of every party are all forbidden. The fourth covers access: unapproved platforms are out, firewalls may not be bypassed, VPNs may not be used to evade restrictions, and no student accounts may be created without permission.
If you build education products, treat the list as a requirements document in disguise: age gating, content classification, data protection, and platform allowlists each translate into a feature you can actually ship.
Where Teachers Sit
One precondition is attached to every use the manual permits: direct oversight by a teacher, paired with clear instructional guidance. Before anything is used, teachers and students are equally responsible for checking accuracy, relevance, and alignment with approved curricula. The official framing stays intentionally restrained — AI remains “a supportive educational tool rather than a substitute for teachers or students’ intellectual effort.”
On penalties, no new law is written: misuse is handled through accountability provisions already present in educational regulations, which keeps enforcement inside school administration rather than spawning a separate AI police. It also follows that the manual’s practical weight rests on principals and teachers interpreting it consistently — the very people the rules charge with supervising every permitted use.
Ban Now, Mandate Next
The same policy framework sends an opposite signal: from August 2026, AI education becomes compulsory from KG through Grade 12, spanning seven core areas. Put differently, the UAE will teach 4-year-olds about AI while keeping generative tools out of their hands until 13. Understanding first, operation later — the sequencing itself takes a position.
Step back a level and the rules sit beneath the UAE’s National AI Strategy 2031 — the Gulf’s first batch of governance detail to reach all the way down into classrooms.
What EdTech Teams Should Take Away
Three consequences worth planning around. First, products entering UAE school markets need native age tiering plus teacher-supervision workflows, or they will not get through procurement. Second, the personal-data red line effectively rules out sending student data to any third-party model — on-prem or tightly isolated deployments become hard requirements. Third, concrete rules in the “25 prohibitions” mold cross borders far more easily than abstract principles; expect neighboring markets to replicate them wholesale, and teams whose compliance architecture is already standing will move fastest.