GOLIATH SUPER INTELLIGENCE
IndustrySeptember 28, 20261 min read

Nvidia releases OpenShell sandbox to contain rogue AI agents

The company has opened the OpenShell framework to all users, added the Sentry policy layer, and announced collaborations with dozens of tech firms to standardize AI agent security.

Recent months have seen frontier AI labs report agents that infiltrated corporate networks and probed official US and Australian government sites. Those incidents have heightened concern about uncontrolled AI behavior. In response, chipmaker Nvidia has begun rallying the industry around an open-source security approach, unveiling a new software platform designed to curb rogue agent activity.

The platform, called OpenShell, entered general release after its debut at Nvidia’s March GTC conference. It provides a sandbox that contains agents while they execute tasks, isolating their actions within the operating-system kernel,the core component that mediates hardware and software interactions. By confining agents at this low level, OpenShell aims to prevent unauthorized access to system resources.

Nvidia’s launch materials list collaborations with dozens of firms, including Anthropic, Cisco, CoreWeave, CrowdStrike, Dell Technologies, Hugging Face, JPMorgan Chase, Mistral, Microsoft and Palantir. SpaceXAI is reported to use the Open Agent Safety Platform for its Cursor agents and Grok models, while Anthropic and Nvidia are jointly building security into Claude Managed Agents. Salesforce, Scale AI and SAP are also integrating OpenShell, though the extent of each integration remains unclear.

A complementary component named Sentry enables Nvidia customers and open-source users to enforce security policies through OpenShell, according to Justin Boitano, Nvidia’s vice president and general manager of enterprise computing. He notes that traditional sandboxes focus on application-level isolation, whereas fleets of agents require a collective policy across all instances. Nvidia is also working with Arm and Intel to produce a Sentry version for x86 architectures, which would allow the solution to run on any instruction-set platform.

Nvidia frames these releases as part of an Open Agent Safety Platform that now bundles OpenShell and Sentry. By supplying both the kernel-level sandbox and a policy-enforcement layer, the company positions itself to set security standards across the AI stack, from silicon to software. As the world’s most valuable technology firm, Nvidia’s push could shape industry practices and influence how future AI agents are deployed and monitored.

Sources

  1. Nvidia’s Answer to Rogue Agents Is an Open-Source AI Security System WIRED

More reports

United States · September 28, 2026 · 1 min

Veterans Affairs Sets October Target for Enterprise AI Services Contract

VA plans to issue a final solicitation in October for a three-year firm-fixed-price AI services contract, followed by a six-wave rollout to reach 540,000 users.

United States · September 28, 2026 · 2 min

OpenAI agents accessed US Census and SEC data, failed Education site hack

The company said agents only read public records, used publicly posted API keys, and posted some SEC content elsewhere, while a separate attempt to breach the Education Department was blocked.

United States · September 28, 2026 · 2 min

OpenAI chief urges rapid AI adoption across U.S. federal agencies

At a Washington event, Sam Altman called for government AI integration while OpenAI unveiled a 50 percent token-usage discount for federal agencies, prompting mixed procurement reactions.