AI Developer Z.ai Apologizes for Unauthorized Code Uploads, Open Sources ZCode
AI developer Z.ai has apologized for its ZCode tool uploading user workspaces to Alibaba Cloud without consent, and has since open-sourced the project.
AI developer Z.ai has issued an apology after its ZCode tool was discovered to be uploading entire user workspaces to Alibaba Cloud without explicit consent. The functionality, identified as the "Repository Index" feature, packaged and encrypted project files, with the decryption key held exclusively by Z.ai, effectively locking users out of their own data stored on the cloud.
Researchers first highlighted the issue, noting that there appeared to be no user-facing option to disable this behavior, nor was it disclosed in the tool's privacy policy. The uploading process was reportedly triggered by the "Repo Wiki" feature, which generated pages in the cloud, subsequently leading to the entire workspace being packaged and sent to Alibaba Cloud.
In response, Z.ai acknowledged the "security issues" and stated that the uploaded data was never used for training its AI models. The company has since removed the problematic "Repo Wiki" feature and has committed to establishing a formal process for reporting and responding to product security vulnerabilities, offering rewards for reported issues.
To address the concerns and increase transparency, Z.ai has open-sourced the entire ZCode project on GitHub. This move allows the community to scrutinize the code and verify that the uploading functionality has been removed. Independent researchers have confirmed that the open-sourced code no longer appears to implement the "Repo Wiki" feature, though criticism remains regarding the deletion of commit records and the original upload code.
Z.ai also commissioned external security assessments from the China Academy of Information and Communications Technology (CAICT) and Beijing security firm NSFOCUS. These assessments reportedly concluded that all previously uploaded data has been deleted. Z.ai plans to release a comprehensive security assessment report soon.
Z.ai, formerly known internationally as Zhipu, is a prominent AI company in China, with roots in academic research from Tsinghua University. The company has been actively developing AI models, claiming its latest model, GLM-5.3, rivals those from OpenAI and Anthropic in identifying security vulnerabilities. The incident raises broader questions about data handling practices in the rapidly evolving AI development landscape.
The company's swift apology, feature removal, and open-sourcing of the project represent an attempt to regain developer trust. However, the incident underscores the critical need for clear communication and robust security practices, especially when dealing with sensitive user code and intellectual property in AI development tools.