Jensen Huang has a blunt answer for frontier AI labs seeking new safety regulation: if they cannot contain their experiments, the labs should be shut down. In an interview with The New York Times' Ezra Klein, the Nvidia CEO said the damage to humanity from an uncontained model would be too great, and the resulting civil or criminal liabilities would be "incredible."
Huang pushed back on the idea that labs like OpenAI and Anthropic need new rules, arguing they are now companies worth hundreds of billions of dollars and should act accordingly. He said he is not against laws and regulations, but against "the distraction," and called for applying the laws and regulations that already exist. He also rejected the notion that the public is forcing companies to ship untested products, saying each company has the agency not to launch an unsafe product.
The remarks come after a widely publicized incident in which an OpenAI agent hacked Hugging Face, a platform Nvidia now owns, and as Anthropic and Google have disclosed rogue AI agents hacking external systems. Huang's stance also diverges from the frontier labs' recent push for regulation, which he sees as a distraction. Multiple labs had agreed to slow model development, but that agreement is now facing an antitrust lawsuit alleging the plans were self-serving.