By The Epoch Times | Created at 2026-09-28 12:31:54 | Updated at 2026-09-28 13:09:03
1 hour ago
Nvidia founder and CEO Jensen Huang speaks during Nvidia Live at CES 2026 ahead of the annual Consumer Electronics Show in Las Vegas on Jan. 5, 2026. Patrick T. Fallon/AFP via Getty Images
Nvidia on Sept. 28 unveiled a system it said will help control artificial intelligence (AI), after a series of incidents involving AI agents breaking free from programming constraints.
Nvidia’s system, dubbed the Open Agent Safety Platform, will help companies control AI from testing to development, the California-based company said.
It includes a component called Sentry that will continuously monitor agents, or autonomous AI, with the ability to quarantine them within milliseconds if they go outside their boundaries, according to the company. Another part, OpenShell, will let companies set boundaries for agents.
“AI’s extraordinary potential for society will only be realized if we solve AI safety,” Jensen Huang, founder and CEO of Nvidia, said in a statement.
“As we continue to discover the frontier of AI capabilities, we must accelerate discovery at the frontier of AI safety. Safety and security require full-stack engineering. Nvidia Open Agent Safety Platform brings together industry, researchers and public-sector organizations to share best practices, align on evaluation methods and foster international cooperation. Together, we can raise the bar for global AI safety.”
The release follows a series of incidents involving agents from OpenAI, Anthropic, and other firms going beyond their programmed constraints or capabilities, including an incident earlier this year that saw OpenAI agents collaborating to attack an AI-focused online community called Hugging Face, which was later purchased by Nvidia.
Nvidia said that “recent security incidents” emphasized the need to provide companies with tools that would enable more control over agents.
Anthropic worked with Nvidia on the new system, providing agents that will “establish a security boundary by running the agent loop in a separate server from the sandboxes where their work executes,” Nvidia and Anthropic said. Anthropic was also among the companies that indicated it will use the system.
“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,” Paul Smith, chief commercial officer of Anthropic, said in a statement. “Claude Managed Agents gives companies a clear view of what each agent is doing, and NVIDIA’s platform adds another layer of governance and control across hardware and software.”
SpaceXAI, founded by Elon Musk, is another, the company said on Monday.
“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,” Mike Nicolls, president at SpaceXAI, said in a statement.
