Nvidia Unveils Open Agent Safety Platform for Responsible AI Development

Nvidia Founder & CEO Jensen Huang spoke to CNBC's 'Squawk Box' on September 28, 2026, about the company's Open Agent Safety Platform and share buyback plan.

Huang emphasized that for AI systems to be delivered safely, they must be designed with restrictive rights. He described this as creating a 'sandbox' or 'force field' around the agent, ensuring it has minimal rights necessary to perform its job.

Huang also discussed the engineering problem of containing and monitoring AI agents, comparing it to the development of web browsers which provide containment systems for applications.

He highlighted that Nvidia's new Open Shell and NVIDIA Sentry technologies make it possible for companies to contain and monitor AI agents. Huang believes that innovation and regulation are not conflicts, but rather a matter of timing: 'at the beginning of the industry, when the technology is being invented and nobody really knows how it's going to turn out, it would be a shame to regulate before understanding.'

The company's stock price jumped 1.5-1.7% on news of the share buyback plan, but Huang considered this less important than the Open Agent Safety Platform.