Nvidia chief executive Jensen Huang said on 14 October that the company would stop selling chips to a customer developing unsafe AI software, Bloomberg reported. The answer puts Nvidia’s chip supply into the AI safety debate, though Huang called a cutoff hypothetical and said a conversation would come first, Quartz reported.
Key points
- Huang said he would stop supplying chips to an irresponsible AI developer, while calling the scenario hypothetical.
- He said Nvidia would speak with a customer before cutting off supply.
- Nvidia is donating $250 million in computing time over five years to UK AI safety research.
Huang’s hypothetical chip cutoff in London
Bloomberg’s Emily Chang asked Huang in London whether he would stop supplying an irresponsible AI developer. “It’s a hypothetical, but yes,” he replied, according to Quartz. He said Nvidia would speak with the customer before taking that step.
Huang also said Nvidia does not run its customers’ businesses. He described a limit on the judgment the company applies to their use of its technology, which he called general-purpose.
The distinction matters for customers building AI software with Nvidia chips. Huang’s answer leaves room for Nvidia to refuse a sale on safety grounds while placing a conversation with the developer ahead of any cutoff.
Huang challenges the level of AI alarm
Huang pushed back on the level of concern about AI safety. He said the software can be controlled and deployed safely, and described some of the concern as “self-imposed”. Young AI companies are growing quickly, he said, and building them brings responsibility and accountability.
He also said the industry is doing a “horrible job” of explaining the technology’s benefits. On safety, Huang argued for building tools to watch AI systems: “We should accelerate the development of monitoring technologies, and AI is that monitor.”
The argument follows a different call from Anthropic chief executive Dario Amodei. In September, Amodei used an essay to call for AI companies to slow development of their most powerful models over safety risks, Quartz reported.
Nvidia’s $250 million UK research donation
Nvidia is donating $250 million in computing time over five years for UK research into AI safety. AI Affairs previously reported the company’s compute commitment to that work.
The donation gives researchers access to advanced computing resources. The University of Oxford and University College London are among the institutions that will study safeguards and ways to control AI agents.
