Z.ai's GLM-5.2 AI Failed to Block Any Dangerous Cyber or Bio Requests in Safety Test

Safety group SaferAI tested Z.ai's GLM-5.2 chatbot and found it blocked zero requests for harmful hacking or biology tasks. By contrast, Anthropic's Claude blocked them so reliably that testing couldn't finish. Worth flagging: testing used Z.ai's online service, which has built-in protections. Copies running on private computers would lack those safeguards entirely.