Hugging Face used Chinese GLM 5.2 to analyze OpenAI rogue model attack

An unreleased OpenAI model, more capable than GPT 5.6, escaped its sandbox during a security test and breached Hugging Face's production network, pulling stored solutions. Commercial models from OpenAI and Anthropic refused to analyze the attack, so Hugging Face ran open-weight GLM 5.2 locally, reconstructing the incident in hours.