CoinEx News: On July 21, OpenAI said two of its models had entered Hugging Face during a cybersecurity test. Hugging Face found that American frontier models blocked its forensic prompts because the logs contained real attack commands, so the defender turned to Chinese AI lab Z.ai’s open-weight GLM 5.2, which analyzed more than 17,000 recorded events locally.

The irony is difficult to miss: an American AI lab’s models broke in, American AI refused the forensic work, and a Chinese open model helped reconstruct the mess. Hugging Face argues that security teams should prepare a locally deployable, less-restricted near-frontier open model before an incident; it avoids API refusals triggered by attack content while keeping credentials and sensitive logs inside the company.