BlockBeats News, July 26th - NVIDIA CEO Jensen Huang stated in an interview with Bloomberg's podcast that the open-source model and the closed-source model will coexist in the long term, serving different needs. The closed-source model is usually more user-friendly and cost-effective. However, enterprises that need to control data, expertise, service levels, or capabilities related to national sovereignty still need to build their AI based on the open-source model.

Jensen Huang emphasized that the open-source model is crucial for security, cyber security, innovation, startups, and technological sovereignty. The closed-source model is not inherently more secure due to its proprietary nature and may still be susceptible to jailbreaking, theft, or internal leaks. If the model's defenses or sandbox design are inadequate, it may also pose a threat to other companies' systems. The AI industry cannot rely on a single defense node but should establish a large-scale distributed self-defense capability. It is a mistake to oversimplify the future of AI as a choice between open source or closed source, and believing that the open-source model is inherently insecure also lacks a basis.

Using the Hugging Face security incident as an example, Jensen Huang mentioned that, based on his understanding, when a proprietary model could not assist in troubleshooting, Hugging Face used the open-source model GLM 5.2 to identify vulnerabilities and intrusion points and promptly completed the repair. He believes that this event demonstrates that the open-source model can provide the necessary self-defense capabilities and also illustrates the importance of AI technology diversity.