Nvidia's Jensen Huang: AI safety and speed can coexist

Nvidia CEO Jensen Huang has emerged as a prominent voice skeptical of calls by some AI leaders to slow model development over safety concerns. AI agents are central to that debate because they can take actions on their own, though regulation of agentic AI specifically gets less attention than broad oversight of the models.

Within Nvidia, VP of agentic AI El Hallak frames the challenge as engineering. He said advances in AI models make "harness engineering" — the layer of instructions and code governing how an AI agent operates in practice — more critical, and warned that a capable agent needs to be built so it can be controlled, not put into a pen it can escape.

El Hallak couldn't comment on whether government regulation of the agentic layer is needed, but pointed to industry initiatives like the Open Secure AI Alliance as a venue for locating where more strengths, guardrails, and safeguards should go. The outcome is likely to hinge on whether such industry efforts can set practical guardrails while, in Huang's words, the world doesn't slow down.