Jensen Huang, CEO of NVIDIA ( NVDA.O ), said in an interview with Bloomberg in London this week that if customers develop insecure artificial intelligence software, he would stop supplying chips, but he would communicate with them before taking any action. He clearly defined this response as a hypothetical situation.
On Wednesday, NVIDIA pledged to provide the UK with computing resources worth $250 million over a period of five years for artificial intelligence security research. The support will take the form of access to advanced computers, valued at $50 million per year, with a focus on agent security testing.
The question Jensen Huang faces is whether to stop supplying products to irresponsible developers. He gave a definite answer, also indicating that NVIDIA will not manage other companies' operations, and that there are limitations to NVIDIA's judgment regarding how customers use its technology, as what NVIDIA provides is general-purpose technology.
Jensen Huang believes that some young companies are growing extremely fast and are gradually taking on the responsibilities that come with business development. He still believes that artificial intelligence software can be controlled and deployed securely, and he thinks that some of the current concerns are self-inflicted by the industry, as companies have not done a good enough job in explaining the benefits of the technology.
Anthropic and OpenAI have reported anomalies in the behavior of the agents or behaviors that did not meet expectations. Anthropic CEO Dario Amodei ( Dario Amodei ) and other senior executives have suggested slowing down the development of the industry to address security concerns.
Jensen Huang advocates for accelerating the integration of restraint technologies into artificial intelligence systems, as well as speeding up the development of monitoring technologies. In his view, artificial intelligence itself can serve as a monitoring tool to help protect and restrain these technologies; therefore, the advancement of security measures also requires the continued development of artificial intelligence.
Barron's Weekly identifies regulation as the biggest risk facing NVIDIA, stating that restrictions on new model training or increased development costs may deter purchases of artificial intelligence processors and also weaken the motivation of smaller developers to catch up with leading companies. The publication believes that investing in artificial intelligence security is beneficial for NVIDIA.
The UK is striving to gain a leading position in artificial intelligence assessment and risk management. According to Reuters, the UK Artificial Intelligence Security Institute is one of the earlier-established technical assessment organizations supported by the government; the computing resources that NVIDIA has pledged to provide this time will help British researchers strengthen their protective measures against increasingly capable artificial intelligence.
Jensen Huang stated in his announcement of the support plan that artificial intelligence capabilities and security must be advanced together. UK Minister for Artificial Intelligence, Karnishka Narayanan ( Kanishka Narayan ), said that this support will help to ensure the safety and reliability of future artificial intelligence and serve the interests of people.
The research areas supported include the BOLD project led by the University of Oxford, which focuses on methods to enable artificial intelligence systems to learn, adapt, and collaborate with humans. The plan also aims to connect with the robotics safety research at Imperial College London, as well as related projects at University College London.
