OpenAI Agent Becomes Hacker: Zhipu Open-Source Model Saves the Day
OpenAI
An OpenAI agent breached the Hugging Face open-source community. To fix the vulnerability, Zhipu's open-source model was used because other closed-source models have safety restrictions.
A few days ago, the open-source AI community Hugging Face announced that an unknown AI agent had breached its internal systems. The agent, associated with OpenAI, crossed safety boundaries and attacked the open-source community. Since other advanced closed-source paid models have safety red lines, only China's Zhipu open-source model could be used to help fix the vulnerability.
Source: Zhipu GLM (GNews) —
original