NVIDIA Launches Open Agent Safety Platform With 100 Industry Partners to Secure Autonomous AI Agents
NVIDIA launched the Open Agent Safety Platform, an open framework with OpenShell runtime and BlueField-based Sentry to enforce policies on autonomous AI agents.
NVIDIA launched the Open Agent Safety Platform, backed by more than 100 organizations, to secure autonomous AI agents operating across models, tools, code execution, and enterprise systems. The platform is built on NVIDIA OpenShell, an Apache 2.0 licensed open-source secure runtime providing kernel-level sandbox isolation and policy enforcement over files, processes, credentials, tools, and network destinations. NVIDIA Sentry extends monitoring and enforcement out-of-band onto BlueField-4 DPUs, placing a control point between agents and their reasoning models. The design targets 'agent drift', where long-running agents deviate from operator instructions across a three-layer application, runtime, and infrastructure model.
- OpenShell is an Apache 2.0 sandboxed runtime enforcing file, process, credential, and network policies
- Sentry enforces policies out-of-band on BlueField-4 DPUs between agent and reasoning model
- Framework addresses 'agent drift' where autonomous agents deviate from operator intent
- Backed by 100+ organizations spanning applications, models, infrastructure, chips, and energy
- Organizations
- NVIDIA
From the original article
3 min read at cybersecuritynews.com
NVIDIA has launched the Open Agent Safety Platform, an open framework designed to secure autonomous AI agents as they gain access to models, tools, code execution environments, data, networks, and enterprise systems. More than 100 organizations across the AI ecosystem back the initiative, including application developers, model providers, infrastructure vendors, chip companies, and energy providers. […]
By Abinaya · cybersecuritynews.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.