On 2 April 2026, ten ministries led by the Ministry of Industry and Information Technology (MIIT) jointly issued the Measures on AI Science-and-Technology Ethics Review and Services (Trial) (Gong Xin Bu Lian Ke [2026] No. 75, dated 30 March 2026). Implementing the State Council’s opinions on deepening the “AI Plus” initiative, the Measures set out the scope, service promotion, responsible bodies, working procedures and supervision of AI ethics review, in order to standardise ethical governance of AI science-and-technology activities.
From principles to procedure
Earlier documents on science-and-technology ethics were largely principled and value-oriented — advancing human wellbeing, fairness and justice, controllability and trustworthiness. Such statements guide companies but are hard to turn into executable steps.
The advance of these Measures is procedural. They break ethics review into operable steps and, reflecting the characteristics of AI activities, specify application and acceptance, general procedure, simplified procedure, expert re-review procedure and emergency procedure. Once procedures are fixed, review no longer depends on individual judgement but has a verifiable record.
For companies, this means ethics review shifts from “writing an explanation” to “following a process” — and the process itself must leave a trace.
What the review focuses on
The Measures state that AI ethics review focuses on human wellbeing, fairness and justice, and controllability and trustworthiness. Reflecting the nature of AI activities, the sources of training data, algorithmic bias, the potential social impact of generated content and the degree of interference with human autonomy all come into view.
This dovetails with TC260’s AI Application Ethics and Security Guideline 1.0, which draws boundaries across development, provision and use — see our analysis of the AI application ethics and security guideline. The combined signal is that ethical requirements are moving from principle down into every stage of development and operation.
Tiered procedures: putting resources where the risk is
The parallel general, simplified, expert re-review and emergency procedures are, in essence, risk tiers. Routine, low-risk activities use the simplified procedure; activities with significant ethical risk or high public attention go to the general or even expert re-review procedure; sudden situations use the emergency procedure.
The practical implication is that a company must first be able to tell which tier its activity falls into. Get it wrong and it either wastes procedural cost or treats a risky activity as a formality. Building an internal risk-tiering standard therefore matters more than simply creating an ethics-review post.
What companies need to do
First, identify the responsible body. Be clear which level of ethics committee handles review, avoiding both “no one accepts applications” and “anyone can review”.
Second, embed review in the R&D process. Ethics review should intervene at project initiation, data collection, model training and launch, not be retrofitted afterwards — consistent with the lifecycle governance approach in our AI law practice.
Third, produce deliverable documents. Applications, conclusions and re-review opinions must all leave a record for regulatory inspection. The document system can be built alongside data-compliance documentation — see our data-compliance service.
Fourth, watch the interface with sector-specific ethics requirements. The education-sector framework is advancing in parallel — see our analysis of the reference framework on AI ethics in education; medical and other sensitive sectors are stricter and must also satisfy the requirements of the sector regulator.
Ethics review is not an obstacle to innovation. For companies genuinely working on frontier models or medical and education applications, a review record that stands up to re-examination is itself evidence of capability that can be shown to clients, investors and regulators.
This article is based on public information concerning the Measures on AI Science-and-Technology Ethics Review and Services (Trial) (Gong Xin Bu Lian Ke [2026] No. 75) issued on 2 April 2026. It is for general reference only and does not constitute legal advice.