This paper addresses a central challenge in China’s Judicial Digital Reform: balancing extensive data sharing with security and controllability in the era of generative artificial intelligence (AI). As China advances judicial digitization, the scale of judicial data sharing has expanded rapidly. However, this progress is accompanied by prominent governance issues, including inconsistent data quality due to non-standardized collection, privacy risks from outdated de-identification, ambiguous distribution of data authority among stakeholders, and inherent AI risks such as opacity, bias accumulation, and insufficient output review. These issues threaten judicial fairness, data integrity, and public trust, thereby obstructing the reform’s core goals of secure and efficient data sharing and intelligent judicial applications. To respond, this paper proposes a comprehensive risk governance framework. It emphasizes: (1) source and process control through unified data standards, dynamic de-identification, and blockchain-based traceability; (2) a “primary–partial” authority framework granting courts ultimate authority, with supervised participation by technology platforms and rights-based access for litigants and the public; and (3) enhanced checks on generative AI via explainable AI, weak-to-strong AI supervision, and intelligent auditing of outputs. This risk control framework aims to achieve a sustainable balance between data sharing and security, thereby providing a systematic solution to advance transparency, fairness, and intelligence in China’s judiciary in the digital era.