Post by Yu-Hsiang Tsai

Inspired by the famous slogan from Taiwan's civic tech community g0v: "Ask not why nobody is doing this. First, acknowledge that you are 'nobody'." After a hellish year and a half, I have officially open-sourced the SovereignAudit: Sovereign AI Governance & Audit Framework on GitHub! This is a plug-in defense layer designed specifically for high-risk and Sovereign AI environments. It completely abandons traditional LLM prompt-engineering defenses. Instead, it utilizes O(1) matrix operations combined with state machine logic, transforming Sovereign AI safety guardrails into computable "Cost Flows" and the "Boundary 4 Questions". 🔗 [https://lnkd.in/gvi2vSYT] I know that building a comprehensive architecture like this is a massive undertaking, and it truly needs the open-source community's help to reach its full potential 🥲. What I could do was give my all to provide a complete conceptual framework, hoping it serves as a solid starting point for everyone's reference and future development. P.S. I am absolutely exhausted Q_Q. This framework is the result of countless sleepless nights. To all the experts reviewing it: please go easy on me! Over the past 1.5 years, I've completely overhauled this architecture at least three times... (spits blood) 有鑑於g0v 的標語：「不要問為何沒有人做這個，先承認你就是『沒有人』」，在經歷地獄的一年半，「SovereignAudit 主權 AI 治理與審計架構」我放在Github開源了~ ​這是專為高風險、主權AI的場域設計的外掛式防護層。屏除傳統 LLM 提示詞防護，改用 O(1) 矩陣運算與狀態機等配合，把主權AI自定義與安全防護轉化為可計算的「代價流向」與「邊界四問」。 ​[https://lnkd.in/gvi2vSYT] 我知道要完成這個架構是一個艱鉅的任務，但它需要大家的幫助才能完成🥲，而我能做的，就是盡力提供一個完整的思路，希望能成為大家參考的起點。 P.S. 我好累 Q_Q 這是爆肝生出來的架構，各路大神審閱請鞭小力一點...這一年半……我大改版至少三次以上阿（吐血）