Z.ai

GLM-5.3 Model Demonstrates Autonomous Cyber Exploitation Capabilities

Anthropic reports that GLM-5.3 can generate end-to-end exploits, highlighting significant limitations in the model's safety protocols.