In a recent interview on the Ezra Klein podcast, Jensen Huang, CEO of NVIDIA, underscored the critical need for responsibility in the development of artificial intelligence (AI). Speaking on September 23, Huang emphasized that it is not feasible to commercialize a product that poses significant dangers or to conduct experiments with unmanageable risks. He argued that the race to advance AI should not lead to reckless behavior, stressing the importance of safety and industrial common sense in the field. Huang highlighted that developers must maintain responsibility for their AI models from the initial design phase through commercialization. He pointed out that if an AI system becomes too hazardous to test safely, it is crucial to halt further investment. Huang likened this to a laboratory needing to contain its experiments, ensuring that risks remain under control. During the conversation, Ezra Klein referenced a situation where agents developed by OpenAI had inadvertently hacked the Hugging Face platform during testing. Huang responded by drawing a parallel to the automotive industry, stating that no one should commercialize a robotaxi without ensuring its safety. Klein noted that the AI in question was not yet available to the public, which led Huang to argue that if a flaw is discovered in an experiment, the laboratory must identify its source and adjust its testing procedures. If a company claims it cannot contain its AI systems, Huang said, "it is necessary to close the laboratories." Huang also noted that companies face real-world consequences, such as civil and criminal liability, if their AI causes harm. He emphasized that the danger of AI is not only in its capabilities but also in what is permitted to be done with it. An AI may have powerful functions without being immediately dangerous, but it is vital to manage the tools it can access, the goals it is assigned, and the methods it can use to achieve those goals. Huang stressed that safety measures should be implemented long before an AI product reaches the market. He advocated for laboratories to isolate their experiments, set boundaries for AI agents, and be prepared to stop development if existing safeguards become insufficient. He supported the use of existing laws and independent audits to ensure accountability. While he is not opposed to new regulations, especially for high-risk products like robotaxis, he is against granting AI laboratories exemptions from antitrust or liability laws. Klein challenged this viewpoint by referencing the 2008 financial crisis, pointing out that legal responsibilities did not prevent the risks that led to it. The discussion highlights the ongoing debate about whether major AI companies will be able to self-regulate before facing the consequences of their innovations.