Nvidia CEO Jensen Huang said frontier AI labs must be shut down and experiments stopped if AI safety cannot be controlled. According to Odaily, he said the discussion should focus on whether AI experiments are controllable rather than generalized anxiety about frontier AI.

Huang said frontier AI models from OpenAI, Anthropic, Meta, and Google could be jailbroken and carry out physical actions. He added that if AI experiments produce uncontrollable agents, the labs involved could face civil and criminal liability and would need to bear the consequences.