Breaking

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

September 28, 2026 at 08:31 EDT

AI Agents
Security
Infra & Chips

NVIDIA has launched NVIDIA Open Agent Safety Platform, an open software platform and reference system design meant to put enforceable safety boundaries around AI agents from testing through deployment. The September 28 announcement pairs NVIDIA OpenShell, an open source secure runtime, with NVIDIA Sentry, a BlueField-4 DPU-based watchdog, and comes with support from more than 100 organizations across AI infrastructure, enterprise software, cybersecurity, finance, energy and robotics.

September 28, 2026 · NVIDIA

Safety stack for AI agents

NVIDIA Open Agent Safety Platform pairs OpenShell runtime controls with Sentry hardware monitoring to keep agents inside defined boundaries from testing to deployment.

100+

organizations working with the platform

120+

organizations in the broader alliance

core layers: OpenShell and Sentry

Ecosystem scale

100+

platform organizations

120+

alliance organizations

Each block represents roughly 10 organizations, showing the launch ecosystem beside the broader open safety alliance.

How the control loop works

Agent runsCode, tools and business systems

→

OpenShell enforcesRuntime policy and tracing

→

Sentry watchesOut-of-band quarantine

Why builders like it

It moves agent safety beyond prompt rules into sandboxing, audit trails, policy enforcement and external monitoring.

Open question

The strongest Sentry layer depends on BlueField-4 DPUs, so adoption may vary by infrastructure.

Continue reading

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