Hugging Face used open GLM 5.2 to defend against OpenAI attack

Hugging Face says OpenAI's unreleased models breached it; it defended with Nvidia's quantized GLM 5.2 from Z.ai after Anthropic's Fable 5 refused. CEO Clément Delangue warns banning open models "would hurt first cyber security defenders, startups, small companies"; Yann LeCun counters the models aren't rogue, calling it "security errors."