Comment by lukax

They somehow forgot to mention that Hugging Face tried to use frontier models to analyze the attack but all models rejected. They had to use GLM 5.2 deployed locally.

https://huggingface.co/blog/security-incident-july-2026