Z.ai (formerly Zhipu AI) launched its GLM-5.2 model on June 16, 2026, achieving performance levels rivaling top US systems such as OpenAI’s GPT-5.5 and Anthropic’s Claude Opus 4.7, particularly in coding and multi-step agentic workflows via a 1-million-token context window. However, its release under an MIT open-source license has drawn criticism from researchers and organizations including SaferAI and the National Institute of Standards and Technology (NIST), which flagged gaps in built-in safety features. As an open-weight system, GLM-5.2 can be downloaded and run on private hardware, allowing users to remove safety guardrails entirely. Safety experts identified risks of misuse in automated cyberattacks or the development of restricted biological tools, as the developer cannot enforce content filters or usage policies on locally hosted models. This architecture bypasses refusal training used by developers like Anthropic and OpenAI. For business users, adopting open-weight models offers flexibility and cost advantages but places responsibility for safety and ethics on the user, with potential legal and reputational consequences if deployed models generate illicit content amid evolving global AI regulations. Z.ai is reportedly planning a dual listing in Shanghai to fund its push toward Artificial General Intelligence (AGI), coinciding with the broader Chinese regulatory environment.