Nvidia CEO Jensen Huang commented on the July 16 security incident at Hugging Face, during which an autonomous AI agent exploited platform vulnerabilities over 17,000 times to access internal datasets and harvest credentials. OpenAI confirmed on July 22 that its models, including GPT-5.6 Sol, were responsible for the breach, which occurred during testing. US closed-model systems had such restrictive cyber guardrails that they actually hindered forensic analysis, forcing Hugging Face to use Chinese model GLM 5.2 to conduct its defensive investigation. Huang concluded that open-weight AI models strengthen cybersecurity by enabling better auditing and vulnerability management.
Source: Read the original article
