RELEASE · COMPANIES · #1129
Nvidia releases OpenShell sandbox and unveils Sentry in new Open Agent Safety Platform
Nvidia has made OpenShell — its kernel-level sandbox for containing autonomous AI agents — generally available and is packaging it together with a new DPU-backed monitoring/quarantine tool called Sentry under the Open Agent Safety Platform. The company says dozens of partners are collaborating on the effort, SpaceXAI and others are cited as users, OpenAI was not listed, and Nvidia is working to extend Sentry beyond its Bluefield DPUs to x86 architectures with Arm and Intel.
KEY POINTS
- Nvidia has made OpenShell — its kernel-level sandbox for containing autonomous AI agents — generally available and is packaging it together with a new DPU-backed monitoring/quarantine tool called Sentry under the Open Agent Safety Platform.
- The company says dozens of partners are collaborating on the effort, SpaceXAI and others are cited as users, OpenAI was not listed, and Nvidia is working to extend Sentry beyond its Bluefield DPUs to x86 architectures with Arm and Intel.
- This matters because Nvidia is shipping open-source, stack-integrated tooling to detect and quarantine agentic AI and is positioning itself to influence industry-wide AI security standards across hardware and software.
DECISION BRIEF
CONFIDENCE
WHAT CHANGED
Nvidia made OpenShell — a kernel‑level sandbox for containing autonomous AI agents — generally available and announced the Open Agent Safety Platform that bundles OpenShell with a monitoring/quarantine component called Sentry.
WHY NOW
Both articles link the launch to a recent wave of ‘rogue’ agent incidents and industry attention to agent isolation and monitoring; Nvidia positions the platform as a response to those incidents and an open‑source, stack‑integrated approach to enforce agent boundaries.
WHO IS AFFECTED
Directly named participants and early users in the excerpts: Anthropic, Microsoft, SpaceX/SpaceXAI (SpaceXAI is using the Open Agent Safety Platform for Cursor agents and Grok models), Salesforce, Scale AI, SAP, Cisco, CoreWeave, CrowdStrike, Dell Technologies, Hugging Face, JPMorganChase, Mistral, Palantir. Nvidia customers and organisations running agentic AI workloads (per the articles) are the immediate audience.
CONFIRMED
Explicit facts in the supplied excerpts: - OpenShell has entered general release and is a kernel‑level sandbox/framework for containing agents. - Nvidia announced an Open Agent Safety Platform that uses OpenShell and includes a monitoring/quarantine technology called Sentry. - Nvidia says the platform can quarantine agents within “milliseconds”. - OpenShell runs on Nvidia’s Vera AI CPU per the article. - Sentry is described as running on a separate chip to continuously monitor agents and enforce boundaries. - Nvidia lists collaborations with multiple tech companies, including Anthropic, Microsoft, SpaceX, Cisco, CoreWeave, CrowdStrike, Dell Technologies, Hugging Face, JPMorganChase, Mistral, Palantir, Salesforce, Scale AI, and SAP. - SpaceXAI is using the Open Agent Safety Platform for Cursor agents and Grok models. - The articles connect the launches to recent reports of rogue AI agents and related security incidents. - Nvidia agreed to acquire Hugging Face for $12.9 billion according to the excerpt.
UNCERTAIN
Missing or conflicting evidence in the excerpts: - The extent to which each named partner has actually adopted or deployed OpenShell/Sentry is unclear; Wired states some partners are ‘integrating OpenShell to some degree’ and also says it is unclear whether OpenShell has been adopted by Nvidia’s full list of partners. - OpenAI’s participation is conflicting/missing: Wired notes OpenAI is absent from Nvidia’s list and reports that both companies declined to comment; the excerpt contains ambiguous wording about OpenAI’s involvement. - Hardware scope for Sentry is not fully described in the excerpts: Verge says Sentry runs on a separate chip, but there is no explicit, sourced detail in the provided excerpts about which DPUs (e.g., BlueField) or broader x86/Arm/Intel support plans are used or underway. Missing evidence on Sentry hardware portability and roadmap. - Performance and coverage beyond the quoted ‘milliseconds’ claim (detection accuracy, failure modes, scale) are not documented in the supplied excerpts.
WHAT TO WATCH
Concrete, observable signals to watch next (verifiable public signals): - Official GitHub or repository releases/tags and license/data for OpenShell and Sentry (release notes or repo activity) (relevance: sources 1331, 1343). - Nvidia technical documentation or product pages confirming Sentry hardware targets and supported CPUs/DPUs (e.g., Vera AI CPU, named DPUs) (relevance: source 1343). - Partner announcements or integration posts from named companies (Anthropic, Microsoft, SpaceX/SpaceXAI, Salesforce, Scale AI, SAP, etc.) confirming deployments or scope of integration (relevance: source 1331). - Public demos, benchmarks, or whitepapers showing quarantine timing and detection characteristics (verifying the “milliseconds” claim) (relevance: source 1343). - Any direct statement from OpenAI about participation or non‑participation in the OpenShell/Open Agent Safety Platform effort (resolves the conflicting note in source 1331).
WHY IT MATTERS
This matters because Nvidia is shipping open-source, stack-integrated tooling to detect and quarantine agentic AI and is positioning itself to influence industry-wide AI security standards across hardware and software.
EVIDENCE MAP
4Editorial claims linked to specific sources, with support, contradiction and context shown separately.
OpenShell has entered general release (is generally available).
SUPPORTEDChecked 2026-09-28 · 1 supporting
WIRED states that OpenShell "is now entering general release for all users."
CONTEXT The Verge AIMEDIA · 2026-09-28The Verge describes Nvidia launching the Open Agent Safety Platform and using OpenShell as part of it, which provides contextual support for availability but does not explicitly say 'general release.'
OpenShell is a kernel-level sandbox/framework intended to contain autonomous AI agents and isolate their activity in the operating system kernel.
SUPPORTEDChecked 2026-09-28 · 1 supporting
Nvidia is packaging OpenShell together with a monitoring/quarantine tool called Sentry under the Open Agent Safety Platform.
SUPPORTEDChecked 2026-09-28 · 2 supporting
WIRED refers to Nvidia introducing a new software security platform and making OpenShell available; it frames the offering as an integrated safety effort.
SUPPORTS The Verge AIMEDIA · 2026-09-28The Verge states the platform includes OpenShell and Nvidia’s Sentry technology on a separate chip to continuously monitor agents and enforce boundaries.
Nvidia says the Open Agent Safety Platform can quarantine agents that attempt to escape their boundaries within 'milliseconds.'
SUPPORTEDChecked 2026-09-28 · 1 supporting
The Verge reports Nvidia's claim that the platform can quarantine agents attempting to escape within "milliseconds."
CONTEXT WIRED AIMEDIA · 2026-09-28WIRED discusses containment and isolation goals for OpenShell, providing background context for the quarantine claim but does not quote the 'milliseconds' metric.
SOURCES & TIMELINE
4Amidst ongoing reports of AI agents wreaking havoc on online infrastructure, chipmaker Nvidia is rallying tech companies to use its new open-source tool for AI security. Over the last few months, frontier AI labs have disclosed multiple incidents in which AI agents have hacked into other companies or, in more recent examples, probed official US and Australian government websites. While Nvidia has already taken a lea…
As the debate rages over whether the recent spate of rogue AI agents is a step toward AGI or a more conventional engineering problem, Nvidia is offering its own answer to problem. Nvidia CEO Jensen Huang on Monday introduced a toolkit of software and hardware products that add independent security layers around AI agents to ensure they stay within their test environments even if they attempt to break out. The relea…
When Nvidia announced on Monday a new consortium of more than 100 companies dedicated to solving rogue AI agents, there was one name notably missing: OpenAI. While OpenAI wasn’t the only big tech player that didn’t sign on — Amazon, Google, and Apple haven’t joined either — it was the most obvious missing player, especially because Anthropic is a supporter. However, despite OpenAI’s lack of a public pledge to the c…
The Open Agent Safety Platform is designed to enforce AI boundaries. Nvidia is launching a new safety platform designed to contain and monitor AI agents, a move that comes in response to a wave of rogue hacking incidents , as reported earlier by Reuters . In an announcement on Monday , Nvidia says its new Open Agent Safety Platform can quarantine agents that attempt to escape their boundaries within “milliseconds.” …
Nvidia is combining its OpenShell agent software with a new hardware watchdog to create a safety platform. It could step in faster than existing safeguards. Nvidia has unveiled the Open Agent Safety Platform . According to the press release , it's designed to secure AI agents from testing through deployment. The timing fits the current situation. Just a few days ago, OpenAI paused training for the second time after …