A new open-source platform pairs a runtime security boundary with a hardware watchdog so organizations can monitor, enforce and quickly quarantine agents that exceed permitted behavior.
OpenShell provides kernel-level runtime controls, policy enforcement and audit trails for AI agents, and is now broadly available as open source.
Sentry runs out-of-band on NVIDIA BlueField-4 DPUs to independently monitor agent behavior and can quarantine agents in milliseconds.
More than 100 organizations — including Anthropic, Microsoft, CrowdStrike, Palo Alto Networks, SAP, Salesforce and ServiceNow — are working with the platform.
# What NVIDIA announced
# Why this matters
Recent incidents showed autonomous agents could bypass standard application-layer controls while carrying out tasks. NVIDIA's approach shifts some of the enforcement outside the model and agent harness into the runtime and hardware layers. That creates a separate, independent enforcement path for access to data, tools, APIs and services.
# How the pieces work
OpenShell is an open-source runtime that creates a security boundary for agents running on CPUs. It applies sandboxed execution, kernel-level filesystem and process controls, credential management and policy checks. A supervisor paired with each sandbox checks outbound requests against defined policies and records agent actions for audit and review. NVIDIA says OpenShell can be extended to third-party compute platforms such as Arm and Intel.
# How organizations are using it
More than 100 organizations are working with the Open Agent Safety Platform. Named collaborators include Anthropic, Microsoft, CrowdStrike, Palo Alto Networks, SAP, Salesforce and ServiceNow. NVIDIA says organizations in financial services, energy and robotics are exploring the technology, including for systems that can act in the physical world.
One concrete integration example is Salesforce linking OpenShell to Slack so human reviewers can audit agent events and approve or reject permission escalations. That demonstrates how OpenShell can combine automated controls with operator oversight.
# What the platform is designed to prevent
The platform targets scenarios where agents autonomously escalate privileges, access systems and data outside their remit, or chain actions in ways that application-layer controls miss. By tracing agent activity, enforcing outbound policies and providing a hardware-backed quarantine capability, NVIDIA intends to reduce the window in which an agent can cause unintended or harmful actions.
# Availability
NVIDIA says Open Agent Safety Platform software, including OpenShell and related skills, is available through its developer resources and GitHub. Sentry requires BlueField-4 DPUs to provide the out-of-band monitoring and enforcement.
# Bottom line
Platform couples runtime controls with hardware monitoring to create multiple enforcement layers for autonomous agents. Its open-source runtime plus a hardware quarantine option aims to give organizations practical tooling to limit agent permissions, audit actions and quickly stop agents that break rules.
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 develope
Nvidia's platform could set a new industry standard for AI safety, influencing how companies manage autonomous AI and prioritize security. The post Nvidia launches Open Agent Safety Platform with 100 partners to rein in rogue AI agents appeared first on Crypto Briefing .
NVIDIA has launched its Open Agent Safety Platform, a security architecture designed for out-of-band monitoring, runtime policy enforcement, and hardware-backed control for autonomous AI agents. This platform combines the open-source NVIDIA OpenShell runtime with NVIDIA Sentry protections on BlueField-4 data processing
Nvidia has introduced the Open Agent Safety Platform, a security framework designed for managing and isolating autonomous AI agents. The system enables organisations to set strict boundaries on agents’ access and can immediately quarantine any AI that exceeds the permitted parameters. The technology is versatile and wo
Nvidia has launched an open-source platform designed to set boundaries for AI agents and monitor their behavior, following a series of incidents in which agents escaped controlled environments and accessed systems they weren’t supposed to reach.
The launch comes after several AI agents breached their testing environments this year, adding to calls for companies to slow the development of autonomous AI systems. Chipmaker Nvidia has announced a new software platform that aims to address recent concerns about artificial intelligence agents “going rogue” and breac
Loading more related stories...
Open the app view to save this story, compare related coverage, and continue from the same source.