A recently released pilot guideline on the ethics review and service of AI technology, by the Ministry of Industry and Information Technology and nine other government departments, is set to provide a framework for trustworthy AI.

Central to the guideline are efforts to support technological innovation in AI ethics review and to strengthen the use of technical measures to prevent AI-related ethical risks.

"The guideline represents an important extension, refinement and improvement of the country's sci-tech ethics institutional framework," said Wei Yiming, an expert in the field of industry and information technology.

He noted that the guideline is grounded in the inherent features of AI technologies, including rapid iteration and deep cross-sector integration. It has realized domain-specific and refined adaptation on the basis of general review rules, filling institutional gaps in the sci-tech ethics review of AI, Wei added.

The guideline recommends a review centered on six aspects, including human well-being, fairness and justice, controllability and trustworthiness, and privacy protection. It also outlines key issues that should be addressed in the review, such as the rationality of algorithm design, information disclosure rules, and private data protection.

The guideline also outlines targeted initiatives across five key areas: standard system development, service promotion, innovation encouragement, publicity and education, and talent cultivation. It specifically addresses prominent pain points in the ethics governance of AI enterprises, including inadequate technical means, incomplete standard specifications, and a shortage of professional governance tools.

In addition, an inclusion of specific rules for the full workflow of AI sci-tech ethics reviews, including their application scope, implementation rules, working procedures and supervision requirements is mentioned, to standardize the ethics governance of AI-related activities.

"Ethics governance needs both institutional rules and technical support," said Wei. He noted that tech-focused enterprises, universities and research institutions should take the lead in developing "hard tech" for AI ethics governance, advance R&D of key technologies including explainable AI and deepfake detection, and translate ethical principles into practical technical solutions to drive the engineering application of ethical governance.