Uploaded June 2026 | Updated September 2026, 2 weeks ago

---

Autonomous AI agents are shifting from a novelty to a core component of production applications. However, giving AI agents the ability to run code, call APIs, and access databases introduces unprecedented security vulnerabilities.

In this talk at Cloudflare Immerse Stockholm 2026, Daniele Molteni (Product Manager at Cloudflare) details how anyone can build and ship autonomous AI agents safely. Learn how Cloudflare’s platform enables a "Secure by Default" posture by sandbox-isolating agent workloads, managing credential secrets safely, and enforcing strict network perimeters right where the code executes.

📍 Recorded live at Sergel Hub, Stockholm at Cloudflare Immerse Stockholm 2026.

👉 Learn more about the event: fieldmarketing.www.cloudflare.com/immersestockholm-2026

🕒 TIMESTAMPS:
00:00 - Introduction: The Rise of Autonomous AI Agents
02:50 - The Security Nightmare of Unconstrained AI Workloads
07:45 - Building "Secure by Default" Agent Architectures
13:10 - Isolating Agent Code with Edge Workers and Sandboxing
18:40 - Managing API Tokens and Outbound Access Safely