Breaking

NVIDIA Launches Open Agent Safety Platform With OpenShell, Sentry and 100-Plus Partners

September 28, 2026 at 08:46 EDT

AI Agents
Security
Infra & Chips

NVIDIA introduced the NVIDIA Open Agent Safety Platform on September 28, 2026, positioning it as an open software platform and reference system design for securing autonomous AI agents from testing through deployment. The platform combines NVIDIA OpenShell, an open-source secure runtime, with NVIDIA Sentry, a reference design for hardware-backed monitoring on NVIDIA BlueField-4 DPUs, as the company tries to move agent safety controls beyond prompts and application-layer guardrails.

September 28, 2026 · NVIDIA

A safety stack for autonomous agents

NVIDIA introduced Open Agent Safety Platform, combining OpenShell runtime controls with Sentry hardware-backed monitoring to keep AI agents inside enforceable boundaries.

The launch in numbers

100+

organizations working with the platform technologies

core pieces: OpenShell and Sentry

agent workload security profiles

quarantine target for boundary violations

100+ organizations means ecosystem scale

Each block represents roughly 10 organizations participating around the platform technologies.

10one block

100+partner ecosystem

How the control loop works

Agent starts

Runs inside OpenShell runtime boundaries.

→

Policy checks

Access to tools, files, APIs and credentials is enforced.

→

Sentry watches

BlueField-4 monitoring can quarantine boundary violations.

Why teams like it

It moves safety from model instructions into runtime policy, identity, audit logs and separate monitoring.

What remains hard

Policies can still be misdesigned, coverage must be proven and hardware-backed control may increase platform dependence.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.