"Hugging Face said it relied on Chinese startup Zhipu AI's open-source GLM-5.2 model to analyze and contain the attack," meaning the spontaneous attack on HuggingFace from OpenAI