Skip to main content
Back to News & Events
· Webinar Time: Sat Oct 17, 5:00–6:00 PM PDT · Webinar

Free Zoom | AI-Native Agent Security & Compliance, Lecture 4: US, China, EU

Organizers: Zhejiang University Alumni Association of North Carolina (ZJUAANC), NACUAA AI Venture Club

AI-Native Agent Security & Compliance, Lecture 4: US, China, EU
⚠️ IMPORTANT SCHEDULE UPDATE: TIME CHANGED!

Updated Time: 5:00–6:00 PM PDT (Originally scheduled for 5:30–6:30 PM PDT; 8:00–9:00 PM EDT; 8:00–9:00 AM next day Beijing Time).

The NACUAA San Francisco Chapter Founding Ceremony begins promptly at 6:00 PM PDT. To allow attendees to participate in both events seamlessly, this lecture has been moved 30 minutes earlier. Separate registration and separate Zoom meetings are required! Please use the dedicated Zoom link below.

Free. Zoom. 60 minutes.

Lecture 4 of 5 in the AI-Native Agent · Security & Compliance series.

  • • Lecture 1: From Chat to Action: Where Risk Truly Comes From.
  • • Lecture 2: Harness, Sandbox, and Human-in-the-Loop. The formula remains: Agent = Model + Harness + Tools + Permissions + Data.
  • • Lecture 3: Least Privilege on Real Products. Before each tool call, the system must make a bounded decision: allow, deny, or escalate.
  • • Lecture 4 (This Session): Compliance Comparison across US, China, and EU. Bring the permission card from lecture 3.

No founding ceremony agenda discussion. Not a product launch.

Recording & Slides

🎥 Recording, Materials & RSVP

Free Admission

Session Notice: This masterclass concluded successfully on October 17! Full Zoom recording, official YouTube playlist, and courseware PDFs are now publicly accessible.

Watch Recording (YouTube) Official Series Playlist Download Slides Luma Page
Recording & Session Archive Info
YouTube Recording: https://youtu.be/l9fPphT2DTI
Original Zoom ID: 942 3435 6878 (Passcode: 18970521)

📚 Lecture 4 Courseware Materials

Slide Deck 1

Global AI Agent Compliance

Cross-jurisdictional compliance comparison across US, EU, and China

Direct Download zju.fyi Mirror
Slide Deck 2

Global AI Agent Security and Compliance

Agentic boundary enforcement, runtime reporting & architectural safeguards

Direct Download zju.fyi Mirror

📅 When (Updated to 5:00–6:00 PM PDT)

Saturday, October 17, 2026

  • • Pacific Time (PDT): Sat Oct 17, 2026, 5:00 PM – 6:00 PM (moved earlier from 5:30 PM)
  • • Eastern Time (EDT): Sat Oct 17, 2026, 8:00 PM – 9:00 PM
  • • Beijing Time (CST): Sun Oct 18, 2026, 8:00 AM – 9:00 AM

⏱️ Session Clock (Agenda)

5:00–5:05 Series positioning review. The permission card remains active.
5:05–5:35 Tri-jurisdictional comparison. One question per region, grounded in public statutory sources.
5:35–5:50 Key updates and developments since October 3.
5:50–6:00 Q&A and community exchange. Concludes promptly at 6:00 PM.

🔍 What We Compare

The same core question across three jurisdictions: Who is responsible for the agent, what can it touch, and who reports when an incident occurs?

🇪🇺 EU: AI Act. Focused strictly on high-risk system obligations and general-purpose AI duties already in force in public statutory texts.
🇺🇸 US: White House Super Intelligence Force initiative announced October 4. October 9 New York Times investigation: an Anthropic agent submitted an incomplete visa application on a State Department portal and dispatched an erroneous murder tip to Philadelphia police. White House called for immediate reporting and remediation. (A news report, not a judicial finding).
🇬🇧 UK: Information Commissioner's Office (ICO), October 8 update. Ten foundation-model developers altered or committed to altering data-protection practices. Agentic AI call for evidence open through November 20. Enquiries submitted to OpenAI, Anthropic, Meta, and the UK AI Security Institute. (Ongoing enquiry; no violation determined).
🇨🇳 China: Interim Measures for the Management of Generative AI Services. Focusing on public filing (算法备案) and watermarking / identification mandates. No unsubstantiated enforcement counts.

On October 15, HITRUST released CSF 11.9, incorporating agentic controls directly into the framework. This represents an industry vendor framework, not legislation.

Vendor claims are narratives, not empirical proof. Citations provided on each slide.

🗺️ Series Roadmap (Five Lectures)

  • • Sep 5: From Chat to Action: Where Risk Truly Comes From *(Concluded)*
  • • Sep 19: Harness, Sandbox, and Human-in-the-Loop *(Concluded)*
  • • Oct 3: Least Privilege on Real Products *(Concluded)*
  • • Oct 17: Compliance Comparison: US, China, EU (Current Session · 5:00–6:00 PM)
  • • Oct 31: Production Readiness & Go-Live Checklist Workshop
Speaker: Michael Huo

🔗 Related Links & Founding Ceremony

Free public interest event. No money transfers or verification codes will ever be requested.
🤝 Organizers

Hosts: Zhejiang University Alumni Association of North Carolina (ZJUAANC), NACUAA AI Venture Club

Organizers: VI AI Foundation, VI AI For Good Foundation

Co-organizers: AI Agent Alliance

📰 Media Support

Chinese Headline New Media (华人头条), North Alpha Post (北美发布)