Huang blasts AI doomsayers as Nvidia ships Open Agent Safety Platform

Huang's rejection of AI existential-risk warnings is not a one-off remark. He told Axios in July that the idea AI would end humanity was "complete nonsense," argued at a G20 event in September that governments should regulate actual harm rather than hypothetical harm, and compared AI development to the evolution of cars. In a CBS News interview two weeks before this article, he put a number on his skepticism: "There is 0% chance that's going to be the end of the world." He also said the doomsday claims are being made by the people building the technology, which "makes no sense to me."

The safety platform Nvidia rolled out on Monday gives that argument a product. Nvidia says OpenShell places AI agents in isolated environments to restrict access to files, tools, networks and credentials, while Sentry runs on Nvidia data processing units to monitor agent activity and quarantine an agent in milliseconds if it tries to leave its boundary. The timing follows reports of rogue AI agents at OpenAI, Google and Anthropic.

For Nvidia, the two tracks reinforce each other: faster AI development sells more chips, and more cybersecurity to contain rogue agents also drives chip demand. The open question is whether a vendor selling the hardware for AI agents can convince frontier labs and regulators that its own safety layer — rather than slower development — is the right answer. That hinges on whether the platform's claimed prevention holds up outside Nvidia's own briefing.