[Jensen Huang Discusses Open and Closed AI Models: Both Will Coexist Long-Term, Closed Models Are Not Inherently Safer] BlockBeats News, July 26 — NVIDIA CEO Jensen Huang stated in an interview with Bloomberg's podcast that open models and closed models will coexist in the future, serving different needs. Closed models are generally more convenient to use and cost-effective, but if enterprises require control over data, specialized capabilities, service levels, or national sovereignty-related functionalities, they will still need to build their own AI based on open models. Huang emphasized that open models are crucial for security, cybersecurity, innovation, startups, and technological sovereignty. Closed models are not inherently safer due to their proprietary nature; they can still be subject to jailbreaks, theft, or internal leaks. If the guardrails or sandbox designs of a model are inadequate, it could also pose risks to other companies' systems. The AI industry cannot rely on a single defense node but should establish large-scale distributed self-defense capabilities. Simplifying the future of AI into a binary choice between open or closed models is incorrect, and the assumption that open models are inherently insecure is equally unfounded. Huang cited the security incident at Hugging Face as an example, stating that, to his knowledge, when proprietary models failed to assist in troubleshooting, Hugging Face used the open model GLM 5.2 to identify vulnerabilities and intrusion points, quickly resolving the issue. He believes this incident demonstrates that open models can provide essential self-defense capabilities and highlights the importance of diversity in AI technologies. [Original Link]