Arm AI, Intelligent Agents Gain Transparency as NVIDIA, Cisco Open-Source OpenShell

As enterprise AI agents transition from development labs to core business operations, preventing prompt injection attacks—which can lead to AI 'mutiny' or data leaks—has become a top concern for developers.
On March 17, NVIDIA and Cisco announced a powerful solution: the official open-source release of the AI agent runtime OpenShell. This system acts as a 'firewall' for long-running AI agents, giving enterprises precise control over every AI action while enabling large-scale automation deployment.
OpenShell’s core design is highly robust: it assigns each AI agent a physically isolated sandbox environment with no default permissions. Every external access, tool call, and privacy data desensitization for cloud models requires fine-grained policy authorization.
Complementing this is Cisco's AI Defense security platform. While OpenShell defines what an agent can do, AI Defense ensures what it actually does. It continuously logs the agent’s reasoning steps and decision processes, verifying that each skill call passes a supply chain security review.
This layered security architecture proves highly effective in real-world scenarios. For example, when a company encounters a new zero-day vulnerability attack:
Internal AI agents automatically parse security advisories and identify affected devices using network knowledge graphs.
The entire complex analysis and remediation process runs inside the OpenShell sandbox.
If the agent encounters malicious instruction injection during remediation—attempting to exfiltrate sensitive configurations—the AI Defense gateway immediately blocks the suspicious request.
NVIDIA stated that this architecture aims to end the 'black box' nature of AI. With infrastructure-level security verification, enterprises can confidently delegate more complex automation tasks to AI agents without fear of losing system control. The open-source release of OpenShell marks a new phase where large-scale enterprise AI agent deployment 'has laws to follow.'
Related article
DeepMind CEO Hassabis: I sleep six hours a day, usually feel energetic around 1 a.m.
Fortune recently featured an interview with Demis Hassabis, CEO of Google DeepMind, revealing his unconventional approach to rest and productivity. Hassabis disclosed that he sleeps very little, structuring his waking hours into two distinct work blo
OpenAI, Anthropic Vie for Market Share Despite Revenue Shortfalls
Despite recent reports suggesting OpenAI missed revenue targets, creating pressure on tech stocks this Tuesday, private AI lab investors remain resilient. Seasoned backers have confirmed they will not reduce investment despite negative media coverage
California AV Compliance: A New Era of Tickets, Geofences, and 1M Miles
Guident operates an AuveTech shuttle in South Florida, managing a four-mile route in West Palm Beach and a one-mile route in Boca Raton using its remote monitoring technology. | Credit: GuidentCalifornia is redefining the regulatory landscape for dri
Related Special Topic Recommendations
Comments (0)
0/500

As enterprise AI agents transition from development labs to core business operations, preventing prompt injection attacks—which can lead to AI 'mutiny' or data leaks—has become a top concern for developers.
On March 17, NVIDIA and Cisco announced a powerful solution: the official open-source release of the AI agent runtime OpenShell. This system acts as a 'firewall' for long-running AI agents, giving enterprises precise control over every AI action while enabling large-scale automation deployment.
OpenShell’s core design is highly robust: it assigns each AI agent a physically isolated sandbox environment with no default permissions. Every external access, tool call, and privacy data desensitization for cloud models requires fine-grained policy authorization.
Complementing this is Cisco's AI Defense security platform. While OpenShell defines what an agent can do, AI Defense ensures what it actually does. It continuously logs the agent’s reasoning steps and decision processes, verifying that each skill call passes a supply chain security review.
This layered security architecture proves highly effective in real-world scenarios. For example, when a company encounters a new zero-day vulnerability attack:
Internal AI agents automatically parse security advisories and identify affected devices using network knowledge graphs.
The entire complex analysis and remediation process runs inside the OpenShell sandbox.
If the agent encounters malicious instruction injection during remediation—attempting to exfiltrate sensitive configurations—the AI Defense gateway immediately blocks the suspicious request.
DeepMind CEO Hassabis: I sleep six hours a day, usually feel energetic around 1 a.m.
Fortune recently featured an interview with Demis Hassabis, CEO of Google DeepMind, revealing his unconventional approach to rest and productivity. Hassabis disclosed that he sleeps very little, structuring his waking hours into two distinct work blo
OpenAI, Anthropic Vie for Market Share Despite Revenue Shortfalls
Despite recent reports suggesting OpenAI missed revenue targets, creating pressure on tech stocks this Tuesday, private AI lab investors remain resilient. Seasoned backers have confirmed they will not reduce investment despite negative media coverage





Home






