On 12 May 2026 the Reference Framework on AI Ethics in Education was released at the World Digital Education Conference held in Hangzhou, Zhejiang. Covering AI use across basic, higher and vocational education, the framework provides a systematic ethical framework and code of conduct and establishes principles such as “the human remains the subject”.

Why education needs a dedicated ethical framework

Education differs from ordinary consumer settings. Its users include many minors; the teacher-student relationship is inherently unequal; and evaluation results directly shape a student’s path. Once AI enters tutoring, guidance, assessment and management, the risks of bias, dependency and manipulation in its output are amplified by the educational relationship.

General AI governance rules can set a floor but struggle to cover education-specific ethical problems. That is the point of a dedicated framework: translating abstract principles into concrete conduct in educational settings. A framework that stays at the level of slogans would not help a school decide what an AI tutor may or may not do with a pupil’s data.

What “the human remains the subject” means

“The human remains the subject” means the responsible party in education is always a person — the teacher, the school, the administrator — and AI is a tool, not a substitute. In practice this has two requirements. First, educational decisions cannot be handed over to algorithms automatically: for key decisions such as assessment, progression and class assignment, human judgement and review must be retained. Second, AI’s role must be clearly labelled, and students have a right to know whether they are interacting with a machine or a person.

For edtech companies, this means product design cannot sell “replacing the teacher” as a feature; it should position AI as an aid and reserve interfaces for human intervention at key points.

Product and compliance requirements for edtech companies

First, positioning and marketing. Products such as intelligent tutoring, Q&A coaching and virtual teaching assistants must align their messaging with an “assistive” positioning, avoiding claims that they replace or outperform a teacher’s judgement.

Second, lawful data processing. Education involves large volumes of minors’ personal information, which is sensitive personal information; processing requires separate guardian consent and the minimum-necessity principle — see AI data compliance risks.

Third, content safety. AI output aimed at students needs content filtering and labelling to avoid generating inappropriate content, consistent with the requirements set out in the Qinglang minors online protection guide.

Fourth, teacher-facing tools are equally constrained. AI systems used for teaching management and teacher evaluation touch on teachers’ personal information and labour-management scenarios, as discussed in the AI employment compliance guide.

The education ethics framework is not an isolated document; it must connect with regimes on minors’ online protection, personal information protection and generative AI governance. Structurally it continues the “ethics first, rules to follow” approach, in line with the earlier AI safety governance framework.

The practical implication for companies is to move ethical requirements into the early stages of product design and data processing rather than patching them in after launch. For support on data compliance in education, minors’ information processing and product compliance assessment, see our Data Compliance service.

This article is based on public information on the release of the Reference Framework on AI Ethics in Education at the World Digital Education Conference on 12 May 2026. It is for reference only and does not constitute legal advice.