Archives open audit has become a mainstream scenario for the application of "artificial intelligence + archives". The integration of artificial intelligence into archives open audit discretion essentially involves enabling machines to participate in discretionary decision-making. This approach not only facilitates the realization of citizens' rights to access archives but also enhances the consistency of archives open audit outcomes. However, the application of artificial intelligence in archives open audit also gives rise to multiple risks, including weakened subject autonomy, evasion of discretionary procedures, distortion of algorithmic discretion, and blurred accountability. In this regard, it is essential to uphold a regulatory philosophy of inclusiveness and prudence, adhere to the principle of guiding technology toward good through the rule of law, and establish a human-in-the-loop AI-assisted archival disclosure review system. A procedural framework for AI-based archives open audit should be shaped through technical due process, while an end-to-end calibration mechanism for discretionary decision-making in AI-assisted archives open audit must be implemented. Furthermore, a comprehensive accountability system with aligned responsibilities and incentives should be established.