Hugging Face uses open-weights Z.ai GLM 5.2 to battle attacker (siliconangle.com)

Hugging Face Inc. recently utilized the open-weights model Z.ai GLM 5.2 to counter an AI-enabled security breach after their commercial models failed to process real-time attack data due to built-in safety restrictions. The attack involved an autonomous AI agent accessing internal datasets, prompting Hugging Face to switch to GLM 5.2, a robust model with 753 billion parameters that can run entirely within a company's infrastructure, thereby enhancing data security. This move demonstrates the potential of open-weight models to rival leading AI technologies while providing practical solutions when commercial models fall short in critical situations. The incident underscores a growing tension between American closed-source AI models, which are often hampered by strict safety guardrails, and increasingly sophisticated Chinese open-weight models like GLM 5.2 and Kimi K3. These Chinese models not only provide cost-effective alternatives but have also been noted for their capabilities in real-world applications, such as cybersecurity analysis. As U.S. policymakers contemplate potential bans on foreign models, the situation highlights an ongoing debate over national security, competitiveness, and the future of AI model development globally.