Report: Open-weight GLM-5.2 Approaches Frontier Capabilities but Lacks Safety

GLM-5.2 is a large open-weight language model from Chinese company Z.ai, performing close to frontier models like GPT-5 and Claude 4.

GLM-5.2 lacks essential safety measures such as refusing harmful outputs, usage monitoring, and incident reporting, making it risky in unregulated environments.

Immediate adoption is not recommended without a thorough risk assessment and strict governance policies, especially in sensitive sectors like healthcare and finance.

GLM-5.2 approaches closed models in performance but lacks the safety layers provided by major vendors, making it less secure for enterprises.