Hugging Face CEO Clement Delangue thanked a Chinese AI model for helping investigate a security breach tied to an OpenAI hack, after American commercial AI refused to assist. The company ran the Chinese model GLM 5.2 locally to probe the incident. Delangue said the episode holds an important lesson.
The Refusal
When Hugging Face needed help analyzing the breach, it turned to American commercial AI providers. Those providers declined to cooperate. The company did not name the firms or give reasons for the refusal. But the rejection forced Hugging Face to look elsewhere.
The Alternative
Hugging Face then ran GLM 5.2, a model from Chinese AI lab Zhipu AI, on its own local infrastructure. That allowed the team to investigate the hack without relying on external commercial services. The CEO publicly thanked the Chinese AI for stepping in. The move highlights a growing willingness to use open-source models from different countries when proprietary options are unavailable.
The CEO's Takeaway
Delangue said the incident carries an important lesson, though he did not elaborate on specifics. The implication is clear: when one door closes, another opens — and sometimes that door leads to a competitor's technology. For Hugging Face, the lesson may be about resilience and the value of keeping multiple AI options ready.
The breach itself involved an OpenAI hack, but Hugging Face did not disclose details about the attack or what data was compromised. The company has not said whether it will change its security protocols or continue using GLM 5.2 for future investigations.
