Uploaded July 2026 | Updated September 2026, 2 weeks ago

---

Adel El Hallak and Kris Murphy from NVIDIA will share how teams can secure long-running AI agents from setup to sandboxing, with a focus on permissions, execution boundaries, and controls for agentic enterprise systems.

Key Takeaways:
Long-running agents require identity, access, sandboxing, and policy controls before production deployment
Tool use and memory increase security requirements across data, applications, and infrastructure
Operational guardrails help teams audit, contain, and safely scale agentic workflows

Industry: All Industries
Topic: Agentic AI and Reasoning AI,Claws & Long Running Agents / Generative AI - Code / Software Generation
Technical Level: Technical - Beginner
Intended Audience: Developer / Engineer
NVIDIA Technology: Blueprint,DGX Spark,DGX Station,NeMo

Explore NVIDIA Cybersecurity AI: nvidia.com/en-us/solutions/ai/cybersecurity
Add guardrails with NVIDIA NeMo Guardrails: docs.nvidia.com/nemo/guardrails/latest

#NVIDIAGTC