NVIDIA launches platform aimed at securing autonomous AI agents

SANTA CLARA, Calif.—NVIDIA has unveiled the NVIDIA Open Agent Safety Platform, designed to help organizations govern and control autonomous artificial intelligence (AI) agents as enterprises increasingly deploy agentic AI across business operations, critical infrastructure and robotics applications. 

The offering combines open-source software and hardware-level security controls intended to monitor agent behavior, enforce policies and prevent AI systems from exceeding authorized permissions. 

The launch addresses a growing concern among enterprises and security teams as AI agents gain the ability to perform complex tasks across applications, data sources and operational systems with increasing levels of autonomy. 

“AI’s extraordinary potential for society will only be realized if we solve AI safety,” said Jensen Huang, founder and CEO of NVIDIA. “As we continue to discover the frontier of AI capabilities, we must accelerate discovery at the frontier of AI safety.” 

According to NVIDIA, recent security incidents involving AI agents have highlighted the need for stronger controls that extend beyond traditional application-level security measures. The company said organizations require tools capable of enforcing boundaries around long-running agents that can access multiple systems and resources. 

The company is positioning the platform as a foundation for secure deployment of AI agents in enterprise and highly regulated environments where governance, auditability and policy enforcement are critical requirements. 

A number of major technology, financial services and cybersecurity organizations are participating in the initiative, including Anthropic, Cisco, CrowdStrike, Dell Technologies, Hewlett Packard Enterprise, Hugging Face, JPMorganChase, Microsoft, Palantir, Palo Alto Networks, Salesforce, SAP, Scale AI and ServiceNow. 

Anthropic said it is integrating the platform with its Claude Managed Agents offering to provide additional guardrails around agent access and behavior. 

“Companies are giving AI agents more of their most important work, and they need to direct and verify what those agents do, especially in sensitive environments,” said Paul Smith, chief commercial officer at Anthropic. 

Financial institutions including Citi and JPMorganChase are collaborating with NVIDIA on open-source agent safety technologies, while energy and infrastructure organizations such as Schneider Electric, Siemens Energy and NextEra Energy are working with the platform for critical infrastructure applications. 

The announcement also extends into robotics, an area that could have implications for physical-world security and safety. NVIDIA said robotics companies including Figure, Gecko Robotics and Skild AI are using OpenShell to embed agent safety controls into autonomous systems operating in physical environments. 

“As customers rely more on agents to get real work done, safety should be enforced outside the model by additional controls the agent can’t get past,” said Mike Nicolls, president of SpaceXAI. 

The company said the Open Agent Safety Platform is intended to serve as an open ecosystem bringing together industry, researchers and public-sector organizations to advance standards, evaluation methods and best practices for AI safety and governance. 

Related: Verkada, NVIDIA partner to advance physical AI capabilities 

Photo courtesy of NVIDIA