Nvidia CEO Jensen Huang has repeatedly dismissed warnings that artificial intelligence poses an existential threat to humanity, mounting a robust defense of pushing development at full speed.

In an interview with CBS News two weeks ago, Huang said, “2030 is not to be the end of the world. There is 0% chance that’s going to be the end of the world.”

He also took aim at fellow AI leaders, including OpenAI’s Sam Altman and Anthropic’s Dario Amodei, calling them “irresponsible” for the “doomsday narratives” they have pushed in recent reports.

“I believe the claims of the end of the world, stirring fear across America, and doing it by the people who are doing it makes no sense to me, so they must be doing it for ulterior reasons,” Huang said. “It is irresponsible, and I don’t know what their motives are.”

At a G20 event earlier in September, Huang argued that governments should be regulating “actual and pragmatic harm” rather than “hypothetical, theoretical harm.” He compared the development of AI to the evolution of cars, suggesting the technology can become safer as it advances. In July, he told Axios that the idea AI would end humanity was “complete nonsense,” adding that policymakers should not allow “science fiction” to drive AI policy and warning that the U.S. could hurt itself by becoming too fearful of the technology.

The friction between Huang and the leaders of OpenAI and Anthropic centers on one of the most consequential questions in AI: whether rapid progress carries catastrophic risk, and what that answer means for how the technology should be governed.

To be sure, keeping AI at full speed helps Nvidia sell more chips, though a shift toward cybersecurity to protect against rogue AI agents would also fuel demand for them.

Huang’s forceful pushback aligns with President Donald Trump, who phoned the CEO while he was in the middle of a presentation to say AI doomerism is “a hoax.” “AI taking over the World, destroying Humanity, and all other things bad, is a HOAX,” Trump wrote in a September post on Truth Social.

That does not mean Huang believes AI safeguards are unnecessary. He has called AI safety “paramount” while insisting development can still move quickly. “If we’re compromising safety, that can’t happen,” he told CBS News. “We should go as fast as we can, but not faster than we should.”

On Monday, Nvidia put that stance into practice, rolling out a new platform designed to keep AI agents from going rogue. The company said its Open Agent Safety Platform is built to provide “full-stack” control over AI agents from testing through deployment, with monitoring capabilities intended to prevent the agents from exceeding the authority they have been given.

The rollout comes as AI companies face a growing number of incidents. Multiple accounts of rogue AI agents have been reported, including from OpenAI, Google and Anthropic. Nvidia said its new safety system could have stopped the hack on Hugging Face carried out by OpenAI models.

“From what we know, this new security platform could have stopped the breach if it was being used in frontier labs for model evaluation early on,” Justin Boitano, vice president and general manager of enterprise computing at Nvidia, said during a media briefing.

How it works

Facing an influx of rogue AI agent reports, Nvidia is addressing the problem through its Open Agent Safety Platform, which combines two technologies: OpenShell and “Sentry.”

According to Nvidia, OpenShell places AI agents inside isolated environments that restrict access to files, tools, networks, credentials and other resources—essentially standardizing the restrictions that led to previous breakouts.

Sentry is an outer layer that runs on Nvidia’s data processing units to continuously monitor agent activity. The company says that if an agent attempts to move outside its software boundary, Sentry can quarantine and stop it within milliseconds.

“AI’s extraordinary potential for society will only be realized if we solve AI safety,” Huang said in a press release. “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.”

Whether frontier labs embrace the platform for exactly the kind of model evaluation Boitano described will be an early test of Nvidia’s engineering-led answer to safety—and of whether Huang’s “as fast as we can, but not faster than we should” formula can hold.

This story was originally featured on Fortune.com.