Anthropic flags weak safety defenses in Z.ai's GLM-5.3

Anthropic warned that GLM-5.3, the latest model from Chinese startup Z.ai, approached Claude Mythos Preview on an exploit-development benchmark but lacks sufficient safety protections. The company raised concerns that malicious actors could abuse the model.

• GLM-5.3 from Z.ai nears Claude Mythos Preview on exploit benchmark
• Anthropic says the model lacks sufficient safety protections
• Concerns grow over potential malicious use of the model