Nvidia CEO Huang: Open and Closed-Source AI Models Will Coexist Long Term; Neither Inherently Safer

According to BlockBeats, Nvidia CEO Jensen Huang said on July 26 during a Bloomberg podcast interview that open and closed-source AI models will coexist long term, serving different needs. Closed-source models offer convenience and lower costs, but enterprises requiring data control, specialized capabilities, service-level agreements, or sovereignty will need to build AI systems based on open models.

Huang stated that closed-source models are not inherently safer due to their proprietary nature and can still face jailbreaks, theft, and internal leaks. He cited Hugging Face's security incident as an example, noting that when proprietary models failed to diagnose the issue, Hugging Face used the open-source GLM model to identify vulnerabilities and complete repairs, demonstrating how open models provide necessary self-defense capabilities.