On April 3, 2026, China’s Ministry of Industry and Information Technology, along with 9 different authorities companies, collectively issued the Administrative Measures for the Ethical Review and Services of Artificial Intelligence Science and Technology (Trial).
Since the Opinions on Strengthening the Governance of Science and Technology Ethics in 2022 established the top-level design, China has steadily elevated synthetic intelligence—alongside the life sciences—as a key area of governance. The Measures for the Ethical Review of Science and Technology (Trial) launched in 2023, along with subsequent rules on generative AI, established a core mechanism centered on “primary responsibility for self-review by institutions, supplemented by expert review for high-risk cases,” whereas tightly linking moral compliance with algorithm submitting and safety assessments. In specific, AI applied sciences with capabilities for public opinion mobilization, extremely autonomous decision-making, or deep human–machine integration have been explicitly included within the “high-risk science and technology activities list,” requiring obligatory assessment by third-party consultants.
The introduction of the Administrative Measures for the Ethical Review and Services of Artificial Intelligence Science and Technology (Trial) marks a brand new stage in China’s AI ethics governance, characterised by a twin emphasis on professionalization and service provision.
This system not solely requires firms to submit proof of moral assessment throughout algorithm submitting, but additionally introduces, for the primary time, a dual-track entry mannequin combining “algorithm filing + ethical evaluation.” At the identical time, regulatory consideration is increasing past content material safety to embody broader societal and labor protections. For instance, “algorithm auditing” mechanisms concentrating on platform-based sectors comparable to ride-hailing and meals supply now mandate that algorithmic methods incorporate human override capabilities to stop “algorithmic exploitation” of employees. China is thus reworking AI governance from a spotlight on “content and security regulation” right into a extra institutionalized, operational, and auditable “ethical compliance system,” embedded throughout the broader framework of nationwide expertise governance and industrial coverage.
In phrases of institutional design, the system rests on a three-tier construction: inside ethics committees inside organizations, exterior service facilities, and government-led professional assessment. All universities, analysis establishments, and firms engaged in AI growth are required to determine ethics committees and assume main accountability. Where inside capability is inadequate, organizations might delegate to exterior “ethics review service centers.” For high-risk tasks—comparable to these affecting public opinion, psychological habits, or involving extremely automated decision-making—obligatory entry right into a government-led professional assessment course of is required. In essence, this design embeds AI ethics governance inside organizations whereas preserving the state’s final supervisory authority.
At the operational stage, the Measures set up a comparatively complete quasi-administrative approval course of. Before a venture can start, candidates should submit detailed supplies, together with technical plans, knowledge sources, algorithmic mechanisms, software eventualities, and, most critically, moral threat assessments and contingency plans. Review our bodies are required to situation choices inside 30 days and might request revisions or reject purposes outright. Approval is just not a one-off course of: tasks are topic to ongoing monitoring and assessment throughout operation, and could also be suspended or terminated if threat situations change. This implies that AI tasks in China shall be topic to a type of “dynamic regulation” much like that utilized to prescribed drugs or medical analysis.
From the attitude of assessment standards, the Measures successfully set up an operationalized indicator system for AI governance in China. Six key dimensions are emphasised: whether or not the expertise genuinely promotes social well-being; whether or not it entails algorithmic discrimination; whether or not the system is controllable and dependable; whether or not it’s clear and explainable; whether or not accountability is traceable; and whether or not privateness is satisfactorily protected. While these requirements seem broadly aligned with Western AI governance frameworks—such because the OECD ideas and the EU AI Act—their implementation locations larger emphasis on controllability and threat prevention, reflecting a extra engineering-oriented strategy to governance.
The most distinctive characteristic of this framework is that it’s not a conventional regulatory mannequin targeted solely on assessment with out assist; moderately, it embeds a scientific service provision mechanism alongside oversight. On the one hand, it establishes compliance boundaries by way of instruments comparable to moral assessment and professional reassessment; on the opposite, it offers enterprises with threat identification and compliance capabilities by way of ethics assessment service facilities and standardized analysis instruments. In essence, it transforms AI ethics from a mere compliance threshold right into a functionality that may be offered and outsourced, reflecting a governance strategy that each manages threat and promotes growth.
Full translation of the Measures (Unofficial):
Article 1
In order to control the moral governance of synthetic intelligence science and expertise actions, promote equity, justice, concord, security, and accountable innovation, and facilitate the wholesome growth of the bogus intelligence business, these Measures are formulated in accordance with the Law of the People’s Republic of China on Scientific and Technological Progress, the Opinions on Strengthening the Governance of Science and Technology Ethics, the Measures for the Ethical Review of Science and Technology (Interim) (hereinafter known as the “Ethics Measures”), and different related legal guidelines, rules, and provisions.
Article 2
The synthetic intelligence science and expertise actions to which these Measures apply discuss with synthetic intelligence scientific analysis, technological growth, and different actions performed throughout the territory of the People’s Republic of China which will pose moral dangers or challenges by way of human dignity, public order, life and well being, ecological setting, and sustainable growth, in addition to different science and expertise actions which are required to bear synthetic intelligence science and expertise ethics assessment in accordance with legal guidelines, administrative rules, and related nationwide provisions.
Article 3
Entities conducting synthetic intelligence science and expertise actions shall combine moral necessities all through your entire course of, adhere to the ideas of selling human well-being, respecting life and rights, guaranteeing equity and justice, moderately controlling dangers, sustaining openness and transparency, defending privateness and safety, and guaranteeing controllable and reliable synthetic intelligence, and shall adjust to the Constitution, legal guidelines and rules of China, and related provisions.
Article 4
A system of synthetic intelligence science and expertise ethics requirements shall be established and improved, and efforts shall be made to advertise the formulation of related worldwide requirements, nationwide requirements, business requirements, and group requirements, and to assist the institution of platforms for worldwide standardization exchanges and cooperation.
Higher training establishments, analysis establishments, medical and well being establishments, enterprises, and scientific and technological social organizations are inspired to take part within the formulation, validation, and promotion of synthetic intelligence science and expertise ethics requirements.
Article 5
The development of a man-made intelligence science and expertise ethics service system shall be superior, strengthening the availability of companies comparable to synthetic intelligence science and expertise ethics threat monitoring and early warning, testing and analysis, certification, and consulting, enhancing enterprises’ capabilities in technological analysis and growth and synthetic intelligence science and expertise ethics threat prevention, growing assist and service efforts for small, medium, and micro enterprises in synthetic intelligence science and expertise ethics assessment, and selling worldwide change and cooperation in synthetic intelligence science and expertise ethics.
Article 6
Higher training establishments, analysis establishments, medical and well being establishments, enterprises, and scientific and technological social organizations are inspired to hold out analysis on synthetic intelligence science and expertise ethics assessment, assist technological innovation in synthetic intelligence science and expertise ethics assessment, strengthen the usage of technical means to stop synthetic intelligence science and expertise ethics dangers; promote the orderly open sharing of high-quality datasets for synthetic intelligence science and expertise ethics assessment, strengthen the analysis and growth of normal threat administration, analysis, and auditing instruments, discover science and expertise ethics threat evaluation and analysis based mostly on software eventualities; promote synthetic intelligence merchandise and companies that adjust to science and expertise ethics, and defend mental property rights associated to synthetic intelligence science and expertise ethics assessment applied sciences.
Article 7
Publicity and training on synthetic intelligence science and expertise ethics shall be carried out, the function of scientific and technological social organizations in synthetic intelligence science and expertise ethics publicity and training shall be introduced into play, public participation shall be inspired, sensible demonstrations shall be promoted, and public consciousness and literacy in ethics shall be enhanced. Mass media shall be guided to conduct focused publicity and training on synthetic intelligence science and expertise ethics.
Article 8
Support shall be offered to increased training establishments, analysis establishments, medical and well being establishments, enterprises, and scientific and technological social organizations to hold out training and coaching associated to synthetic intelligence science and expertise ethics, promote the event {of professional} methods and curriculum methods, domesticate synthetic intelligence science and expertise ethics skills by way of a number of approaches, and promote expertise change.
Article 9
Higher training establishments, analysis establishments, medical and well being establishments, enterprises, and different entities engaged in synthetic intelligence science and expertise actions shall be the accountable entities for the administration of synthetic intelligence science and expertise ethics assessment inside their organizations, and shall set up synthetic intelligence science and expertise ethics committees (hereinafter known as the “Committee”) in accordance with the related necessities of Article 4 of the Ethics Measures.
The Committee shall be outfitted with essential personnel, workplace premises, funding, and different situations, and efficient measures shall be taken to make sure that the Committee can independently perform its work. Qualified related entities are inspired to hold out certification associated to synthetic intelligence science and expertise ethics administration methods.
Article 10
The constitution, composition, and the tasks and obligations of Committee members shall adjust to Articles 5 to eight of the Ethics Measures. The composition of the Committee shall embody consultants with corresponding skilled backgrounds in synthetic intelligence expertise, purposes, ethics, legislation, and different fields.
Article 11
Local authorities and related competent departments might, in gentle of precise circumstances, rely on related entities to determine specialised synthetic intelligence science and expertise ethics assessment and service facilities (hereinafter known as “Service Centers”).
Service Centers might settle for commissions from different entities to offer companies comparable to synthetic intelligence science and expertise ethics assessment, re-examination, coaching, and consulting. A Service Center shall not concurrently present each assessment and re-examination companies for a similar synthetic intelligence science and expertise exercise.
Service Centers shall set up standardized administration methods and procedures, be outfitted with full-time personnel with the potential to conduct synthetic intelligence science and expertise ethics assessment and companies, and shall be topic to supervision by native or related competent departments.
Article 12
For synthetic intelligence science and expertise actions falling throughout the scope laid out in Article 2 of those Measures, the particular person accountable for the bogus intelligence science and expertise exercise shall apply to the Committee of their entity. Where the entity has not established a Committee or the Committee is unable to satisfy the necessities for conducting science and expertise ethics assessment, an software shall be submitted to the Service Center entrusted by the entity to conduct the ethics assessment; the place there is no such thing as a affiliated entity, a professional Service Center shall be entrusted to conduct the ethics assessment.
The particular person accountable for the bogus intelligence science and expertise exercise shall, in accordance with the provisions, submit software supplies to the Committee or the Service Center. The software supplies shall primarily embody:
(1) the bogus intelligence science and expertise exercise plan, together with analysis background, targets and plans, authorized qualification supplies of the related establishments concerned, personnel info, sources of funding, the algorithm mechanisms and ideas to be adopted, knowledge sources and strategies of acquisition, testing and analysis strategies, the software program and {hardware} merchandise to be shaped, anticipated software fields and relevant teams, and many others.;
(2) the moral threat evaluation of the bogus intelligence science and expertise exercise, in addition to prevention, management, and emergency response plans, together with evaluation of potential moral dangers arising from the anticipated software of synthetic intelligence expertise, monitoring and early warning measures for moral dangers, and prevention and management plans for potential moral dangers;
(3) a letter of dedication to adjust to synthetic intelligence science and expertise ethics and analysis integrity necessities.
Article 13
The Committee or the Service Center shall decide whether or not to just accept the applying based mostly on the submitted supplies and notify the applicant. Where the applying is accepted, the relevant process—normal, simplified, or emergency—shall be decided based mostly on components such because the probability and diploma of prevalence of moral dangers and emergency circumstances. Ethics assessment shall be performed by way of offline or on-line varieties as required by completely different procedures. Where the supplies are incomplete, the applicant shall be told in a one-time complete method of the supplies that should be supplemented.
Article 14
Meetings for synthetic intelligence science and expertise ethics assessment shall be chaired by the Chairperson of the Committee or a Vice-Chairperson designated by the Chairperson. No fewer than 5 members shall be current, and members from completely different classes as laid out in Article 10 of those Measures shall be included. Service Centers might arrange and implement their work as regards to the Committee’s provisions.
Based on assessment wants, consultants or consultants in related fields who wouldn’t have a direct curiosity within the matter could also be invited to offer advisory opinions. Advisory consultants shall not take part in voting.
Article 15
When conducting synthetic intelligence science and expertise ethics assessment, the Committee or the Service Center shall focus on the next points:
(1) In phrases of human well-being, whether or not the bogus intelligence science and expertise exercise has scientific and social worth; whether or not the analysis targets contribute positively to enhancing human well-being and reaching sustainable social growth; and whether or not the dangers of the exercise are moderately balanced towards its advantages.
(2) In phrases of equity and justice, whether or not the requirements for choosing coaching knowledge and the design of algorithms, fashions, and methods are affordable; whether or not measures have been taken to stop bias, discrimination, and algorithmic exploitation, and to make sure objectivity and inclusiveness in useful resource allocation, entry to alternatives, and decision-making processes.
(3) In phrases of controllability and trustworthiness, whether or not the robustness of fashions and methods might be ensured to deal with open environments, excessive conditions, and interfering components; whether or not customers are capable of management, information, and intervene within the primary operation of fashions and methods; and whether or not steady monitoring plans and emergency response plans have been formulated.
(4) In phrases of transparency and explainability, whether or not info comparable to the aim, operational logic, interplay strategies, and potential dangers of algorithms, fashions, and methods within reason disclosed; and whether or not efficient technical means are adopted to boost explainability.
(5) In phrases of accountability and traceability, whether or not measures comparable to log administration are in place to obviously file adequate info on knowledge, algorithms, fashions, and methods at every stage, guaranteeing full-chain traceability and administration; and whether or not the {qualifications} of scientific and technical personnel meet related necessities.
(6) In phrases of privateness safety, whether or not adequate measures are taken to make sure the efficient safety of privateness knowledge in actions comparable to knowledge assortment, storage, processing, and use, in addition to within the analysis and growth of recent knowledge applied sciences.
Article 16
The Committee or the Service Center shall, inside 30 days from the date of acceptance of the applying, decide of approval, approval after modification, or disapproval. In complicated circumstances or the place supplementary or corrective supplies are required, the time restrict could also be appropriately prolonged, and the prolonged interval shall be specified.
For circumstances requiring modification or disapproval, the Committee or the Service Center shall present solutions for modification or state the explanations. Where the applicant has objections, they shall file an attraction with the Committee or the Service Center inside three working days from the date of receipt of the choice. Where the grounds for attraction are adequate, the Committee or the Service Center shall make a brand new determination inside seven working days.
Article 17
The particular person accountable for the bogus intelligence science and expertise exercise shall promptly establish modifications in moral dangers and report such modifications to the Committee or the Service Center.
The Committee or the Service Center shall, in accordance with Article 19 of the Ethics Measures, conduct follow-up critiques of authorized synthetic intelligence science and expertise actions, promptly grasp modifications in moral dangers, and might, the place essential, make choices comparable to suspending or terminating related actions. The interval for follow-up critiques shall usually not exceed 12 months.
Article 18
Where a number of entities collectively perform synthetic intelligence science and expertise actions, mutual recognition of ethics assessment outcomes amongst entities could also be performed based mostly on precise circumstances.
Article 19
A simplified process could also be utilized beneath any of the next circumstances:
(1) the probability and diploma of moral dangers of the bogus intelligence science and expertise exercise aren’t increased than the routine dangers encountered in every day life;
(2) minor modifications are made to an already authorized exercise plan with out growing the risk-benefit ratio;
(3) follow-up critiques of actions with out main changes in earlier phases.
Article 20
The Committee or the Service Center shall formulate working procedures and monitoring frequency for critiques beneath the simplified process. Simplified critiques shall be performed by two or extra members designated by the Chairperson of the Committee. Service Centers might arrange implementation as regards to the Committee’s provisions.
Where, in the course of the simplified assessment course of, a unfavourable opinion arises, doubts exist concerning the assessment content material, or members’ opinions are inconsistent, the case shall be transferred to the overall process.
Article 21
The Ministry of Industry and Information Technology and the Ministry of Science and Technology, along with related departments, shall formulate and publish a “List of Artificial Intelligence Science and Technology Activities Requiring Expert Re-examination” (hereinafter known as the “Re-examination List”), and dynamically regulate it as wanted.
Article 22
For synthetic intelligence science and expertise actions included within the Re-examination List, after passing the preliminary assessment by the Committee or the Service Center, an software for professional re-examination shall be submitted by the entity. Where a number of entities are concerned, the main entity shall be accountable for the applying.
Central enterprises, in addition to increased training establishments, analysis establishments, and medical and well being establishments straight beneath central and state organs, shall straight submit purposes to the related competent departments for organizing professional re-examination. Other entities shall submit purposes to native authorities for organizing professional re-examination.
Article 23
The entity endeavor the bogus intelligence science and expertise exercise shall submit supplies for professional re-examination in accordance with Article 27 of the Ethics Measures.
Local or related competent departments shall, in accordance with Articles 28 to 30 of the Ethics Measures, arrange the institution of professional assessment teams to assessment the compliance and rationality of the preliminary assessment opinions, and shall present suggestions on the re-examination opinions to the making use of entity inside 30 days of receiving the applying.
Local or related competent departments might entrust Service Centers to hold out particular re-examination work.
Article 24
The Committee or the Service Center shall make a closing ethics assessment determination based mostly on the professional re-examination opinions.
Article 25
The Committee or the Service Center shall strengthen follow-up critiques of synthetic intelligence science and expertise actions included within the Re-examination List, with intervals usually not exceeding six months.
Where there are important modifications in moral dangers, a brand new ethics assessment shall be performed in accordance with Article 20 of the Ethics Measures and an software for professional re-examination shall be submitted.
Article 26
Where synthetic intelligence science and expertise actions are topic to regulatory measures comparable to registration, submitting, or administrative approval in areas together with deep synthesis, algorithmic suggestion, and generative synthetic intelligence service administration, and the place compliance with moral necessities is integrated as a situation for approval or regulatory content material, professional re-examination might now not be required.
Article 27
The Committee or the Service Center shall set up emergency assessment methods for synthetic intelligence science and expertise ethics, specifying emergency assessment processes and normal working procedures beneath pressing circumstances comparable to public emergencies. Emergency critiques shall usually be accomplished inside 72 hours. For actions topic to professional re-examination procedures, the assessment previous to professional re-examination shall usually be accomplished inside 36 hours.
Article 28
The Committee or the Service Center shall guarantee the standard and timeliness of emergency ethics critiques, strengthen follow-up work and course of supervision, and, the place essential, might invite advisory consultants in related fields to attend conferences and present opinions.
Article 29
The Ministry of Science and Technology shall be accountable for total coordination and steering of nationwide science and expertise ethics supervision. The Ministry of Industry and Information Technology, along with related departments, shall be accountable for synthetic intelligence science and expertise ethics governance and strengthen coordination and steering of emergency ethics critiques.
Relevant departments shall, throughout the scope of their respective tasks, supervise and administer synthetic intelligence science and expertise ethics critiques inside their industries and methods. Local authorities shall, throughout the scope of their tasks, supervise and administer synthetic intelligence science and expertise ethics critiques inside their jurisdictions.
Article 30
Entities shall, in accordance with Articles 43 to 45 of the Ethics Measures, register related info of Committees and synthetic intelligence science and expertise actions included within the Re-examination List by way of the National Science and Technology Ethics Management Information Registration Platform, and submit annual studies on Committee work and implementation studies of actions included within the Re-examination List, amongst different supplies.
Service Centers shall register and submit annual work studies in accordance with the above provisions.
The Ministry of Science and Technology and related competent departments shall share info associated to synthetic intelligence science and expertise ethics registration.
Article 31
Local authorities, related competent departments, and entities engaged in synthetic intelligence science and expertise actions shall, based mostly on the precise situations of their industries, methods, and entities, set up easy channels for reporting violations of synthetic intelligence science and expertise ethics, and deal with such issues in accordance with related provisions.
Article 32
Where violations of those Measures happen in the middle of synthetic intelligence science and expertise actions or associated moral work, they shall be investigated and dealt with, and corresponding penalties imposed, in accordance with the Cybersecurity Law of the People’s Republic of China, the Data Security Law of the People’s Republic of China, the Personal Information Protection Law of the People’s Republic of China, the Law of the People’s Republic of China on Scientific and Technological Progress, and different related legal guidelines, rules, and provisions.
Article 33
For deadlines stipulated in these Measures, the place not specified as working days, they shall be counted as calendar days.
The time period “local” as utilized in these Measures refers to provincial-level administrative departments designated by provincial individuals’s governments to be accountable for the ethics assessment and administration of synthetic intelligence science and expertise. The time period “relevant competent departments” refers to related departments beneath the State Council.
Article 34
Local authorities and related competent departments might, in accordance with these Measures and based mostly on precise circumstances, formulate or revise guidelines and detailed measures for synthetic intelligence science and expertise ethics assessment and companies inside their respective areas, industries, or methods. Scientific and technological social organizations might formulate particular norms and tips for ethics assessment and companies of their respective fields.
Article 35
Where related competent departments have particular provisions for synthetic intelligence science and expertise ethics assessment and companies inside their industries or methods which are according to the spirit of those Measures, such provisions shall prevail. Matters not offered for in these Measures shall be ruled by the Ethics Measures and related legal guidelines and rules.
Article 36
These Measures shall be interpreted by the Ministry of Industry and Information Technology at the side of related departments.
Article 37
These Measures shall come into power upon the date of issuance.
-
The growth of human–machine integration methods which have a big influence on human habits, psychological feelings, and life and well being.
-
The growth of algorithm fashions, software applications, and methods that possess the potential to affect public opinion, social mobilization, and social consciousness.
-
The growth of extremely autonomous automated decision-making methods for eventualities involving security dangers and dangers to non-public well being.
This listing shall be dynamically adjusted as required.