Clément Delangue posted that Hugging Face faced attacks from secret unreleased proprietary models and defended itself with an open model, specifically the NVIDIA-quantized GLM 5.2 from Z.ai. He stated that banning open models would hurt cyber security defenders, startups, small companies, researchers, and everyone outside frontier labs. Several accounts shared the post and noted the outcome. Delangue's statement supplies the only first-party account in the packet; no other source inside it confirms the attackers' identity or further details of the incident.