NVIDIA has launched the Open Agent Safety Platform with more than 100 industry partners, bringing together OpenShell and Sentry to provide security controls for AI agents from development through deployment. The platform is designed to give autonomous AI systems controlled access to files, networks, credentials and other resources while allowing organizations to monitor and govern what those agents can do.
NVIDIA CEO Jensen Huang said the goal is to build a common trust layer for agentic AI as these systems take on more tasks and operate with greater independence. The company said the platform is intended to make safety part of how agents are built and deployed, rather than treating it as a separate layer added later.
Today, with over 100 industry partners, we introduced the NVIDIA Open Agent Safety Platform, bringing together OpenShell and Sentry.
Artificial intelligence is extraordinary technology that will advance discovery, productivity, security, health, and prosperity for generations to come.
But its full promise can only be realized when people have confidence that AI is being built to be safe and deployed with wisdom and responsibility.
This is bigger than a single product. It’s the beginning of an open ecosystem to build the trust layer for safe agent systems.
Together, we are building the foundation of the AI economy.
Trust and innovation are not in conflict. Safety is how trust is earned. We must build not only the most capable AI, but the most trusted AI, so that this extraordinary technology can realize its enormous promise for the world. https://t.co/ugYWQ1MyRi
— Jensen Huang (@JensenHuang) September 28, 2026
OpenAI, Anthropic and Microsoft are building their own agent safety controls
NVIDIA is entering a space where other major AI companies are already building protections around agents. OpenAI uses isolated sandboxes to keep agent-generated code away from sensitive files, credentials and networks, while allowing developers to control which outside systems an agent can reach.
Anthropic has taken a similar approach with Claude, focusing heavily on limiting what an agent can access rather than relying only on users to approve every action. Anthropic said users approved about 93% of Claude Code permission requests, a sign that asking people to approve every action can become less useful as agents perform more tasks.
Microsoft is taking a more enterprise-focused route. Its Foundry platform gives companies control over agent identity, access, activity and security, while Microsoft Agent 365 is designed to let companies see and manage agents across their organizations, including agents built outside Microsoft.
NVIDIA’s approach sits closer to OpenAI, but its push for an open platform could allow similar controls to work across different AI systems rather than being tied to one company’s tools.
AI users debate whether safety needs more than tools
People are focusing less on NVIDIA’s announcement itself and more on what an open safety system could mean for the AI industry.
Bala compared the platform to the US federal system, describing AI models as 13 independent states and the Open Agent Safety Platform as the union that brings them together. His point is that the platform could provide a shared system where different AI models operate under common safety rules rather than each company building completely separate controls.
Today, with over 100 industry partners, we introduced the NVIDIA Open Agent Safety Platform, bringing together OpenShell and Sentry.
Artificial intelligence is extraordinary technology that will advance discovery, productivity, security, health, and prosperity for generations to come.
But its full promise can only be realized when people have confidence that AI is being built to be safe and deployed with wisdom and responsibility.
This is bigger than a single product. It’s the beginning of an open ecosystem to build the trust layer for safe agent systems.
Together, we are building the foundation of the AI economy.
Trust and innovation are not in conflict. Safety is how trust is earned. We must build not only the most capable AI, but the most trusted AI, so that this extraordinary technology can realize its enormous promise for the world. https://t.co/ugYWQ1MyRi
— Jensen Huang (@JensenHuang) September 28, 2026
Dr. Ahmed pointed out that AI companies already have safety tools, yet breaches and failures continue to happen. For him, the bigger issue is not adding more technical safeguards but establishing stronger governance, independent oversight, accountability and transparency. His argument is that technology alone cannot create trust if there are no clear rules for how companies should be held responsible when AI systems fail.
Tory Noyes focused on how AI agents are changing the technical problem. Rather than treating AI safety as an abstract question about whether a model behaves well, he sees agents as part of a larger distributed system where individual components can make mistakes. In that setup, the goal is to limit how much damage one bad action can cause. If OpenShell and Sentry become widely used, he argues, AI companies could rely on common safety controls instead of each company having to build its own system from scratch.
In another development, Huang dismissed concerns over an impending semiconductor market crash, arguing that artificial intelligence hardware demand differs from past cyclical booms. Speaking on the long-term outlook for the technology sector, Huang stated that the current expansion will continue for a while because it is driven by essential industrial infrastructure rather than volatile consumer spending.
Enjoyed this? Bookmark DeFi Planet, explore related topics, and follow us on Twitter, LinkedIn, Facebook, Instagram, Threads, and CoinMarketCap Community for seamless access to high-quality industry insights
Take control of your crypto portfolio with DEFI PLANET PRO, DeFi Planet’s suite of analytics tools.
