NVIDIA'S HUANG: BUILD AN UNSAFE AI AND I'LL CUT OFF YOUR CHIPS

Asked in a London interview whether he would halt chip supplies to an irresponsible AI developer, Nvidia chief executive Jensen Huang answered directly: it was a hypothetical, but yes, he would, after first discussing concerns with the customer. He stressed the limits of Nvidia's role, saying the company does not run other people's companies and that its general-purpose technology can be used in many different ways, and he set out no definition of unsafe AI, threshold or formal review process.

Alongside the warning, Nvidia announced on October 13 a five-year commitment worth $250 million in computing access, $50 million a year, for British AI safety and security research. The access, delivered through UK-based cloud partners, will support the wider safety community plus named groups including Oxford-led BOLD, UCL-led SOFAIR and Imperial College London, working on agents, open models and robotics safety.