Jensen Huang Opposes Slowing Down AI, Yet Supports Third-Party Audits of Frontier Labs

Beating AI News Flash: Jensen Huang opposed Anthropic CEO Dario Amodei's call to slow down frontier AI development at the All-In Summit, but he agreed with one of Dario's proposals—allowing independent third parties into frontier labs to conduct safety evaluations, preferably with multiple institutions participating at the same time to prevent evaluations from losing their independence.

Huang believes that the most dangerous AI safety issues will first emerge in frontier labs, because only these companies have enough compute to push model capabilities to the limit. He believes labs should first get sandboxes, monitoring, and evaluations right, while also accepting external scrutiny similar to financial audits. Companies that control the most powerful AI should also be held to higher standards.

Halfway through the interview, Trump suddenly called Huang. Huang put it on speakerphone directly. Trump called AI taking over the world a "scam," and Huang immediately agreed, saying "You're right." Trump also said "Robots will not take over the world" and "Whoever wins AI wins." Finally, he asked Huang whether he thought the same way he did, and Huang replied, "Yes, Mr. President."

Like Trump, Huang opposes slowing down and does not agree with doomsday narratives about AI going out of control, but on the issue of introducing external audits into frontier labs, he and Dario rarely found themselves on the same side.