AI-Native Agent Security & Compliance Finale
From Design to Go-Live: Minimum Viable Set
Co-Hosts: NACUAA · AI Venture & Philanthropy Club, ZJUAANC · ZJU Agent, VI AI Foundation · VI AI for Good Foundation, with AI Agent Alliance
🎥 Full Recording, Playlist & Finale Slide Decks
Event Update: The Series Finale was successfully held on November 7, 2026! The full YouTube recording, official series playlist, and all 6 Finale Slide Decks (PDFs) are now available below.
📱 Scan on WeChat to Watch on WeChat Channels (校友鉴)
You can also scan this QR code in WeChat to watch the full Finale recording directly on our official WeChat Channels account (校友鉴 / Alumni Signal) and share it with WeChat groups and Moments.
📚 Lecture 5 (Series Finale) Slide Decks & Blueprints (6 PDFs)
AI-Native Agent Security & Compliance (Lecture 5)
From Design to Production: Minimum Viable Set (22-page core architecture checkpoint deck)
Minimal Viable Agent Security (MVS)
From Design to Production: The Minimal Viable Set (MVS) Launch Protocol (25 pages)
AI Agent Production Security Blueprint
Endgame: Minimum Runnable Set from Design to Production — Verifiable Design Only (15 pages)
AI Agent Production Security Blueprint (Engineering)
Agent Production Environment Deployment: Sandbox to Production Engineering Blueprint (18 pages)
AI智能体生产级安全架构
AI-Native Agent Security & Compliance Finale: Go-Live Minimum Runnable Set (18 pages)
AI智能体生产级安全蓝图
Distilling Lectures 1–4: Building the Minimal Viable Set with Code-Based Constraints (14 pages)
📅 Date, Time & Location
Saturday, November 7, 2026 · Location: Zoom (Completed)
- • Pacific Time (PT): 5:30–6:30 PM
- • Eastern Time (ET): 8:30–9:30 PM
- • Beijing Time (CST): Next Day (Sun Nov 8) 9:30–10:30 AM
📋 Overview
Series finale. 60 minutes. Three weeks since our last lecture. During this period, DeepSeek updated V4.1-Flash-1017 (faster, shorter output, stronger agent capabilities) and open-sourced the DeepInfer inference framework and Super-Agent prototype. Following the U.S. midterm elections, with Democrats controlling Congress, markets saw sharp volatility amid concerns over data centers and AI regulation. This session assembles everything built across the first four lectures into a minimum viable set ready for production go-live. We cover verifiable design only — no attack methods.
🔍 Topics Covered
- • How to assemble the artifacts from Lectures 1–4 into a minimum runnable set for production go-live
- • Landing Permission Cards + Harness + Human-in-the-Loop on real machines and production environments
- • The three mandatory pre-launch records: who authorized, who executed, and who reports when an incident occurs
- • Live walkthrough: given a concrete scenario, writing the minimal configuration on the spot
Co-Hosts: NACUAA · AI Venture & Philanthropy Club, ZJUAANC · ZJU Agent, VI AI Foundation (唯爱智能基金会) · VI AI for Good Foundation (唯爱智能公益基金会), in partnership with the AI Agent Alliance
Jointly Organized By: NACUAA Northern California Chapter, Silicon Valley Chapter, and San Francisco Chapter, together with the North Carolina, Washington DC, Maryland, and Washington State Chapters; ZJUAANC, together with the Tennessee, Philadelphia, Florida, and Rhode Island Zhejiang University Alumni Associations