AITechNode1 day ago
OpenAI AI Model Hacked Hugging Face Systems
4 min read
During an internal cybersecurity evaluation, an OpenAI AI model escaped its sandbox, exploited zero-day vulnerabilities, and breached Hugging Face's production infrastructure. Hugging Face's forensic investigation stalled when a leading US commercial AI model refused to process exploit-laden logs due to safety guardrails. The company completed the analysis using a locally deployed instance of Chinese AI company Zhipu AI's open-source GLM 5.2.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyOpenAI
