Generative artificial intelligence embedded in digital government′s data security risks and legal regulation
Zou Huancong, Wang Qing
College of Law and Politics, Nanjing Tech University
Abstract： Generative AI′s deep integration into digital government is significant, but its full-cycle data security risks cannot be ignored. Risks in the four stages of "collection-processing-storage-output" include: collection security/quality issues, processing-stage algorithmic bias/black boxes, storage leakage/accountability gaps, and output distortion/infringement. Drawing on EU "regulate-first" and US "technology-driven" approaches, a phased framework is proposed: classification/source review for collection, ethical audits/list management for processing, soft-hard law collaboration/public-private risk sharing for storage, and sandbox regulation/digital rights confirmation for output. This promotes digital government development and balances technology, risk governance, and legal safeguards.
Key words : generative AI; digital government; data security; administrative rule of law; legal regulation
