Nvidia founder and CEO Jensen Huang made remarkable statements about artificial intelligence security and industry regulations at Salesforce’s Dreamforce conference. According to Huang, artificial intelligence is a complex computing system consisting of hardware and software designed by humans, rather than a different type of intelligence that develops outside human control. For this reason, he argues that the solution to security should come from engineering studies, not new legal regulations. Nvidia CEO believes existing laws and market mechanisms may be sufficient to ensure companies develop safe products. This approach reveals the perspective of one of the industry’s most influential managers on the subject, at a time when discussions continue on how artificial intelligence should be controlled, especially in the USA and Europe.
In his speech, Huang said that if a company is not sure of the functions, capabilities or security of the product it has developed, it should not release it. Stating that companies can ensure that the product is safe enough by slowing down the development speed when necessary, the manager thinks that the market will also put pressure on manufacturers in this direction. In his opinion, there is no necessary choice between speed of innovation and developing safe products, and companies can provide both at the same time. However, he acknowledges that if a product is deemed out of control or not safe enough, companies should slow down the process. For these reasons, Huang concludes that new laws or regulations specific to the field of artificial intelligence are not necessary.
Jensen Huang sees AI security as an engineering problem
Nvidia’s position in the artificial intelligence industry is also behind Huang’s approach. The company’s GPUs are among the key components of the data center infrastructure used to train and run large AI models. In addition, Nvidia is not limited to hardware only, but also offers artificial intelligence models, software tools and various platforms for developers. Huang points out that the basis of this technology is still information processing systems designed by humans and working within certain rules. This perspective differs markedly from the approach of some security researchers, who argue that artificial intelligence could develop into a completely different and difficult to control form of intelligence.
However, whether market pressure alone can ensure product safety has been a long-debated issue in the technology industry. Even if software companies carry out extensive testing, errors can occur that lead to serious consequences. One recent example of this was when a faulty update released by CrowdStrike in 2024 affected Windows systems worldwide. The update caused disruptions in many services, from airlines to banks and different corporate systems, causing thousands of flights to be disrupted. This example can be used in regulatory discussions as one of the cases that show that even when software manufacturers have good intentions regarding security, unexpected consequences cannot be completely eliminated.
Similarly, in artificial intelligence systems, looking only at the initial intentions of developers may not be enough. Various events, research and lawsuits have been brought to the agenda in different areas, from the security vulnerabilities of the models to their possible effects on mental health. The source text also highlights security incidents related to OpenAI models and lawsuits filed against the company following the deaths of young people who communicated with chatbots for a long time. How to determine legal liability in such cases continues to be one of the main discussion topics of artificial intelligence regulations. The extent to which existing product liability and consumer protection laws cover damages caused by artificial intelligence may become clearer with the jurisprudence in the courts.
One option that Huang did not mention is that the industry creates its own security standards. Having AI companies establish common testing methods, model evaluation standards, and security protocols could provide a different oversight mechanism than direct government regulation. In his statements at the All-In Summit, Microsoft CEO Satya Nadella pointed out that artificial intelligence security is not an issue that concerns only the USA. Nadella stated that countries developing artificial intelligence, including China, will face similar problems such as cyber attacks and citizens’ safe use of technology. Therefore, the debate consists not only of whether a new law should be enacted or not, but also what security standards can be applied on an international scale.
For Nvidia, the acceleration of artificial intelligence investments directly supports the company’s growth. In his speech at Dreamforce, Huang stated that his company’s goals have become larger due to the increase in productivity provided by artificial intelligence, and argued that the technology has a wide potential in terms of sectors and countries. In addition, Huang, who supports open-oriented models, has long expressed the opinion that they can strengthen competition against artificial intelligence laboratories that develop closed systems. Huang’s views are also discussed in the context of commercial interests, as the scope of the new regulations may directly or indirectly affect companies that provide artificial intelligence infrastructure, such as Nvidia. Whether artificial intelligence security should be managed only by companies’ engineering processes and market pressure, or whether special legal rules should also be added to these, continues to be one of the main disagreements in technology policies.