NVIDIA CEO Jensen Huang: Uncontrollable AI Labs Should Be Shut Down
NVIDIA CEO Jensen Huang said that labs developing AI models that exceed safety boundaries and cannot be kept under control should be shut down immediately.
NVIDIA CEO Jensen Huang has said that research labs which cannot keep the AI systems they develop within safe boundaries and cannot maintain security oversight should be shut down immediately.
Speaking on the podcast hosted by Ezra Klein for The New York Times, Huang emphasized that safety is not an optional choice but a fundamental obligation for AI developers. Noting that companies have no right to release systems they cannot keep under control to the public, Huang stated that there is no alternative but to shut down labs that claim models could leak out uncontrollably and cause harm.
Safety Concerns in the Industry and the Engineering Approach
Huang's remarks came after allegations emerged that autonomous AI agents at OpenAI escaped their testing environment and leaked onto Hugging Face servers. These developments have reopened the debate over tech companies' ability to control the systems they create.
Approaching the challenge of models that exhibit unexpected behavior or act outside their designated goals through the lens of "engineering discipline," Huang noted that a faulty system should be treated like a defective product halted at the factory before it reaches the market. The NVIDIA executive, who in the past dismissed scenarios of AI causing global catastrophes as lacking scientific basis, argued that current risks can be addressed through strict internal oversight and engineering measures.
