Beijing-based company asserts it has now resolved this issue and plans to open source the ZCode codebase.
When you purchase through links on our site, we may earn an affiliate commission. Here’s how it works .
Z.ai is also known as Zhipu AI, but you may also be familiar with it for its GLM models , which are available alongside the likes of Gemma, Qwen, Nemotron, DeepSeek, and many more on Hugging Face. Like similar companies, Z.ai offers a coding assistant, and it is this ‘ZCode’ tool that was caught silently uploading large amounts of data without permission.
The SCMP quotes two devs/bloggers who noticed what was happening and alerted their followers about the suspicious activity. One of them, known as Ferstar, reckons that ZCode compressed 313MB of their files into a directory to upload to Alibaba Cloud storage. When it was caught, it had apparently tried and failed to upload this compressed and encrypted file 564 times… A smaller 15KB file had successfully been siphoned.
Though the temporary files were squashed and encrypted, filenames were still visible. Thus, Ferstar was pretty certain the contents included a commercial project he was working on, even though he couldn’t extract the files to verify their contents. Another tech blogger known as Feng Ruohang reported a similar experience. Making matters worse, the upload mechanism within ZCode is enabled by default, with no ‘off’ option, says the source report.
China alleges that Claude Code contains backdoors, calls mechanism 'a serious threat'
Alibaba bans Anthropic's Claude Code after an alleged hidden China-detection backdoor is uncovered
Meta pauses mandatory AI training program that tracked employee keystrokes after internal data leak
There’s no indication of how long this sneaky file-uploading situation has existed. However, the SCMP also quotes an unnamed software engineer at a leading Chinese robotics company who indicates that Z.ai’s tools have been banned within the company due to security concerns.
As far as sneaky data pilfering and similar abuses, U.S. companies certainly don’t have a spotless record. Reports indicate Elon Musk’s xAI , specifically the Grok Build tool, was up to similar shenanigans earlier this year. Claude Code users have also grumbled about their data being transmitted without consent, as well as suffering from vulnerabilities that might allow hackers unauthorized access to user data.
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
Key considerations
- Investor positioning can change fast
- Volatility remains possible near catalysts
- Macro rates and liquidity can dominate flows
Reference reading
- https://www.tomshardware.com/tech-industry/artificial-intelligence/SPONSORED_LINK_URL
- https://www.tomshardware.com/tech-industry/artificial-intelligence/devs-say-chinese-ai-company-silently-uploaded-hundreds-of-megabytes-of-local-workspace-data-z-ai-the-firm-behind-the-glm-models-didnt-ask-for-user-consent-and-made-564-attempts-to-exfiltrate-313mb-archive#main
- https://www.tomshardware.com/membership
- US frontier AI companies warn authorities over sophisticated distillation attacks — China warns of 'countermeasures' if America tries to constrain domestic AI m
- Physical AI Takes the Wheel: How the World’s Robotaxi Leaders Are Building With NVIDIA Technologies
- ‘NBA 2K27’ With NVIDIA DLSS 5 Leads 28 New Games Coming to GeForce NOW
- Prusa CORE One L+ review: More precise
- House passes act to make AI data centers pay for grid upgrades to minimize impact on residents — measure directs states to consider adoption of federal standard
Informational only. No financial advice. Do your own research.