# AI Agents Can Now Have Security That Can't Be Compromised By Themselves

- slug: ai-agents-can-now-have-security-that-cant-be-compromised-by-themselves
- date: 2026-03-23
- category: Agentics

According to NVIDIA's technical documentation, OpenShell enforces agent constraints at the environment level — outside the agent process — making policy bypass structurally impossible rather than just discouraged.

---