Huang Says Nvidia Could Cut Chips to Customers Building Unsafe AI Software

Nvidia could stop supplying chips to customers that develop unsafe AI software, CEO Jensen Huang said, describing a hypothetical response as the company expands support for AI safety research in the United Kingdom.

Huang said Nvidia would speak with a customer before taking such a step. He made the comments in a London interview. He described Nvidia's products as general-purpose technology and said the company has limits in assessing how customers use them.

Nvidia pledged computing resources valued at $250 million for UK AI safety research over five years, averaging about $50 million annually. Projects receiving support include the University of Oxford-led BOLD project, as well as research at Imperial College London and University College London. Testing AI agents is among the priorities for the work.

Huang also called for adding safety constraints and monitoring to AI systems. He said AI capabilities and safety should advance together.