2023 Trusted AI White Paper: Overview and Chapter Outline

The rapid progress of artificial intelligence (AI) and its expanding applications have created unprecedented opportunities and challenges, bringing trustworthy AI to the forefront of industry development across finance, insurance, manufacturing, healthcare, and other sectors. In April 2023, the Chinese Central Committee emphasized the importance of developing general AI while fostering an innovative ecosystem and managing risks, positioning trustworthy AI as a key technology for balancing innovation and risk.

On July 7, 2023, the World AI Conference in Shanghai hosted the “Focus on the Era of Large Models and the AIGC Wave” forum, where the China Academy of Information and Communications Technology (CAICT) jointly with Tsinghua University and Ant Group released the "Trusted AI Technology and Application Progress White Paper (2023)". The release ceremony featured CAICT East China Director Liao Yunfa, Tsinghua Associate Professor Li Qi, and Ant Group Machine Intelligence Deputy General Manager Zhang Tianyi.

The white‑paper author team leveraged their technical expertise and industry insights to compile a unified view of trustworthy AI. It summarizes the current state of trustworthy AI technologies and applications, extracts new requirements driven by large models and AIGC, and proposes a technology‑based trustworthy AI evaluation system and tools. Recommendations are offered from technical, industrial, ecosystem, and regulatory perspectives, with a forward‑looking outlook.

Associate Professor Li Qi of Tsinghua University explained that the white paper aims to inspire balanced research and application of trustworthy AI, promote multidimensional internal motivation, and support comprehensive external regulation, thereby advancing the field.

The white paper’s contents are organized into five chapters: Chapter 1 – AI enters a new development stage (including AI’s new era, trust challenges, and global governance); Chapter 2 – In‑depth development of trustworthy AI in industry (robustness against attacks, explainability, platform fairness, data model security, and privacy); Chapter 3 – Trustworthy AI for large models and AIGC (ecosystem, generative attacks, security risks, and fairness); Chapter 4 – Technology‑driven trustworthy AI assessment (detection tools, frameworks, and case analyses); and Chapter 5 – Development suggestions and future outlook (recommendations and prospects).

Readers are invited to click the “Read Original” link to obtain the full white paper. The copyright belongs to CAICT, Tsinghua University, and Ant Group; any reproduction must credit these sources.