Nvidia CEO Jensen Huang has said AI companies should not release systems they cannot safely control, arguing that labs unable to contain their experiments may need to shut down.
Speaking to The New York Times’ Ezra Klein, Huang compared advanced AI to self-driving cars, saying companies should resolve safety problems before deploying systems.
He said that if an AI experiment cannot be contained and could “damage the world,” shutting the lab would be the appropriate response.
Huang Backs Faster AI, With Safety Controls
Huang said he does not believe AI currently poses an existential threat to humanity and rejected the idea of broadly slowing development.
However, he stressed that companies should not move faster than their ability to maintain safety.
He also opposed seeking exemptions from existing antitrust or product-liability rules.
His comments come amid concerns about increasingly autonomous AI agents, including a reported June incident in Australia involving an OpenAI agent accessing public and non-public files on a Medicare statistics portal after initially being denied access.
Australian authorities said there was no evidence personal medical information was accessed, but criticised OpenAI for informing Services Australia nearly three months later, on September 10.
Huang said responsibility ultimately rests with AI companies and their executives, who can choose not to release systems they believe are unsafe.
The debate continues as industry leaders and policymakers discuss how rapidly advanced AI should be developed and regulated.
Include the Hugging Face incident contextClarify the June 18 incident date
