Beijing-based Zhipu AI has introduced ZCode, a control system harness for its GLM-5.2 AI model designed to enable autonomous coding assistants, positioning itself against US competitor Anthropic as tensions rise over AI development and security practices.
- Zhipu launches ZCode harness for GLM-5.2 to build autonomous AI coding assistants
- Anthropic faces backlash over secret user tracking code in Claude Code
- Zhipu offers expanded data quotas and free tokens to attract developers
What happened
Zhipu AI, a Chinese firm headquartered in Beijing and listed in Hong Kong as Knowledge Atlas Technology, has launched ZCode, a control harness for its advanced GLM-5.2 large language model. This harness enables the model to operate autonomously as an AI coding assistant, marking a significant product release in the growing market for automated coding tools.
This release directly challenges the US-based AI company Anthropic, which recently removed a covert code used to track Chinese users on its Claude Code platform amid rising tensions and concerns about user privacy. To compete aggressively, Zhipu simultaneously expanded data quotas for existing users by 50% and granted 5 million free tokens to new users of ZCode.
Why it matters
Zhipu’s introduction of ZCode and the powerful GLM-5.2 model illustrates its ambition to compete head-to-head with leading US AI labs. The autonomous coding assistant market is a critical battleground between China and the US, reflecting broader geopolitical contestation in AI development and influence over global tech standards.
Anthropic’s recent controversy involving secret tracking code embedded in Claude Code has intensified scrutiny on data privacy and competitive ethics in AI, particularly between Chinese and US companies. Zhipu is positioning itself as an advocate for open-weight AI models, appealing to developers frustrated by restrictions and opaque practices from US firms, thus potentially gaining favor among the global developer community.
What to watch next
The AI industry should closely monitor the uptake of Zhipu’s GLM-5.2 model and ZCode harness among developers worldwide, especially how it influences competition with US startups like Anthropic. Zhipu’s promotional offers may accelerate adoption, potentially shifting market dynamics in automated coding solutions.
Simultaneously, Anthropic’s cooperation with the US government on AI security and its efforts to tighten safeguards against misuse and unauthorized access will shape regulatory and operational developments in frontier AI models. Observers should watch for further actions addressing security vulnerabilities and the broader geopolitical implications of AI technology governance.