Five ministries set out an AI-in-education agenda spanning curricula, applications and governance
Notice of the Ministry of Education and four other departments on Issuing the Action Plan for “AI Plus Education”

English translation
Five government departments issued an action plan that sets tasks for AI courses in primary and secondary schools, teachers’ ability to use AI, pilot bases for education applications and safety governance. The 2030 targets are policy aims, not evidence that classroom outcomes have already been validated.
The national policy text covers curricula, applications and governance and can be read alongside classroom studies evaluating AI-generated instructional materials.
Sources and editorial checks
Editorial check:
- Scope
- Research/policy subjects in China
Limitations
- The action plan sets tasks and targets; it does not provide an independent evaluation showing that the tasks have been achieved or learning outcomes have improved.
Editorial analysis
Original analysis based on the sources listed. Research findings, policy goals, and editorial judgment are identified separately; original rights remain with their sources.
Once AI enters the education policy framework, the key question is how to verify its value in classrooms
Sources and timing: The Action Plan for “AI Plus Education,” issued by the Ministry of Education and four other departments, published on April 10, 2026, and dated April 2. This site reviewed the publicly available full text on November 9, 2026. The analysis below is original to this site and is intended as background reading across reporting periods.
The five-department action plan places AI in education within a shared framework for curricula, teachers, use cases, infrastructure and governance. For readers, the key shift is that policy no longer focuses only on exposing students to a particular tool. It also addresses the competencies students should develop, how schools should choose applications, how teachers should participate in design, and who is responsible for data and safety. The plan sets targets for 2030. That is the policy horizon; it does not mean classroom results have already been achieved in 2026.
The document calls for continued development of guidance on AI literacy in primary and secondary schools, support for local curriculum systems, and clear objectives, content and instructional time for different education stages. Two curriculum questions are easy to conflate. First, learning how AI works is different from using AI to complete assignments in other subjects. Second, students’ ability to use a tool and their ability to judge whether its output is reliable require different assessments. If schools count only the number of courses offered or frequency of tool use, they cannot determine whether students understand model errors, bias, privacy or intellectual property boundaries.
For teachers, the plan calls for stronger application skills, resource development and research into teaching models. Teachers do spend substantial time preparing lessons, giving feedback and adapting materials. But fast automated generation does not mean the material fits the instructional objective. Classroom materials must be accurate and appropriate to students’ age, subject, culture and learning needs. The policy gives schools room to explore; whether it reduces workload or improves learning quality still has to be answered through comparison and follow-up in specific settings. Login counts after training, the number of generated materials and the number of demonstration lessons describe adoption; none alone proves effectiveness.
The document also proposes pilot bases for education applications, systems for evaluating intelligent applications and priority use cases. Pilots create opportunities to examine tools under real school conditions, including documenting failures and cases where a tool is unsuitable. At a minimum, an evaluation should state in advance the target students, users, existing teaching approach, comparison design and observation period. For example, if a tool claims to improve reading, teacher satisfaction alone is not enough. If it claims to save teachers’ time, saved time should not be converted directly into student achievement. Different objectives require different outcome measures; otherwise product demonstrations can be mistaken for evidence of educational effect.
The action plan covers profiles of students’ knowledge, abilities and qualities, personalized resource recommendations and education governance data. These systems may make information easier to organize, while also linking minors’ learning histories, assessments and recommendations. The policy discusses ethical standards and safety oversight, but implementation still needs to answer which data must be collected, who can access it, how long it is retained, how errors can be corrected, and how schools and families are informed. Without these conditions, personalized recommendations cannot be judged as educational progress on technical performance alone. The same product’s results in pilot schools cannot be generalized to all schools, especially where devices, internet access and teacher support vary by region.
This document is connected in the policy hierarchy to the 2026 Education Development 15th Five-Year Plan: the AI action plan assigns more specific tasks, while the five-year plan places digitalization, resource allocation and talent development within a broader agenda. Both are worth reading as policy direction, but neither is an experimental report on a classroom intervention. Researchers can turn the plan’s aims for curricula, teachers, applications and governance into testable questions. Schools can first define their most pressing teaching problem, then choose tools and monitoring measures. Families can ask whether schools explain the purpose, data protections and teacher oversight rather than focusing only on whether a product is marketed as “intelligent.”
This site has reviewed the publicly available policy text but has not verified curriculum implementation in each locality, procurement, student outcomes or independent evaluations. The next step is to connect policy provisions separately with specific implementation, original research and cases of failure. The pace of policy change merits tracking; the pace at which evidence develops requires a different timetable. Keeping these timetables distinct makes it possible to recognize institutional change without claiming learning improvements that have not yet occurred.
Source: Ministry of Education of the People’s Republic of China · moe.gov.cn