Nvidia launches OpenShell software and Sentry system for AI safety

by@wallstengine

Summary

Nvidia has launched the OpenShell software and the Sentry reference system design to help companies enforce strict limits on AI agents' access and actions, enhancing security controls independent of the agents themselves. OpenShell serves as an open-source software layer that safeguards access to files and services while working seamlessly with existing AI models, and Sentry incorporates dedicated hardware for real-time monitoring of agent activities. This initiative addresses the need for infrastructure-level controls in enterprise AI, building on the collaboration with notable partners like Anthropic and Salesforce to integrate safety measures into their products. With over 100 organizations, including Microsoft and Cisco, involved, Nvidia's focus is on ensuring that AI agents operate securely within defined boundaries.

Tokens

$NVDA

Analysis

SAP: SAP delivers enterprise resource planning and business software solutions. It is embedding the new platform into Joule Studio for enhanced AI agent controls. Cisco: Cisco develops networking, security, and collaboration technologies for enterprises. It is participating in the deployment of Nvidia’s AI safety platform across its solutions. Nvidia: Nvidia develops GPUs, processors, and AI infrastructure used across computing, data centers, and emerging applications. In this news, the company is launching OpenShell software and the Sentry reference system to provide external controls and hardware monitoring for AI agents. Jensen Huang emphasized the need for full-stack engineering in AI safety and security. Palantir: Palantir builds data analytics and AI platforms for government and commercial clients. It is working with Nvidia to apply the new safety controls in its AI deployments. SpaceXAI: SpaceXAI develops AI models and tools including those used in coding applications. It is adopting Nvidia’s platform to secure Cursor coding agents and Grok models. Anthropic: Anthropic builds large language models and AI systems focused on safety and alignment. It has collaborated with Nvidia to integrate protections into Claude Managed Agents as part of the new platform launch. Microsoft: Microsoft offers cloud computing, software, and AI services through Azure and related products. It is among the organizations working with Nvidia’s OpenShell and Sentry technologies. Salesforce: Salesforce provides cloud-based customer relationship management and enterprise software. It has integrated OpenShell with Slack to add security layers for AI interactions in its ecosystem. ServiceNow: ServiceNow offers workflow automation and IT service management platforms. It is incorporating Nvidia’s OpenShell and Sentry for secure AI agent operations. CrowdStrike: CrowdStrike provides cybersecurity solutions including threat detection and response platforms. It is collaborating on Nvidia’s initiative to secure AI agent environments. Jensen Huang: Jensen Huang is the founder and CEO of Nvidia. He highlighted that safety and security in AI require full-stack engineering in comments on the new platform launch. Palo Alto Networks: Palo Alto Networks specializes in network security, cloud security, and AI-driven threat prevention. It is integrating with Nvidia’s technologies to address risks in autonomous AI systems. AI Agent Security: Enterprise AI agents operating over extended periods require infrastructure-level controls beyond model training to enforce action boundaries. Hardware Monitoring: Dedicated processors enable independent, real-time oversight of AI behaviors separate from the agent’s own environment. Industry Collaboration: Major technology providers across software, cybersecurity, and enterprise platforms are integrating external safety layers into their AI offerings.

Categories

techaiai_agentsmachine_learning
View Original Tweet