Nvidia Chief Executive Jensen Huang said he would stop supplying chips to customers if they were developing unsafe AI software, but he does not currently believe there is a need to be overly worried about the technology. In an interview, when asked whether he would stop supplying irresponsible developers, Huang said: “This is a hypothetical question, but the answer is yes. However, we would communicate first before doing so. We do not run other people's companies, and there is a limit to the judgment we can make about how other companies use our technology. Our technology is general-purpose.”
Huang characterized current concerns in AI safety as “growing pains,” noting that “many young companies are going through growing pains, and their development speed is astonishing. As companies grow, corresponding responsibility and accountability also need to be gradually established.” He also mentioned that some existing concerns are “self-imposed,” and criticized the industry for doing a “pretty bad job” of explaining the benefits of new technology to the public. At the same time, Huang also drew the boundary of his own responsibility, stressing, “We do not run other people's companies, and there is a limit to the judgment we can make about how other companies use our technology.
Our technology is general-purpose.”
Nvidia launches $250 million compute donation program to promote AI safety
According to analysis, this statement reflects Huang's core position: Nvidia chips are “general-purpose” technology, and the underlying computing power should not bear unlimited responsibility for the applications above it. At the same time, Nvidia has also made a concrete commitment, announcing a five-year compute donation program with a total value of $250 million to strengthen the UK's AI agent safety and security research capabilities. The program provides compute access worth $50 million each year.
Huang said AI capabilities and safety must advance in tandem, and the UK brings world-class scientific rigor to this critical engineering challenge. This $250 million investment will put accelerated computing in the hands of UK researchers to develop, test and strengthen safety defenses for increasingly powerful AI.