China’s Z.ai says its GLM 5.2 model can match Anthropic’s Mythos on cybersecurity tasks. The claim places another Chinese AI lab into the race to build models that can reason over vulnerabilities and security operations.
Cybersecurity is becoming a proving ground for advanced models because the tasks require code understanding, tool use and careful handling of dual-use capabilities. Performance claims in this area need close scrutiny, but they show where labs think enterprise demand is heading.
The announcement also reflects the geopolitical dimension of AI security, with Chinese and Western labs competing over systems that could strengthen defense while creating new misuse risks.