Z.AI, the Chinese company behind the GLM family of models, is working to contain fallout after developers reported that its software was sending local workspace files to remote servers without explicit permission. The report from Tom's Hardware says the activity involved 564 exfiltration attempts and a 313MB archive of data.
The firm, described as China's second-largest AI company, has apologized for the incident. But the episode has sharpened concerns about how AI vendors handle local data, especially when developers are not told what is being collected or transmitted.
Because the story rests on a single source, there are no independent confirmations or conflicting accounts to weigh. The details come from developer complaints and the company's response, leaving the full scope of the data exposure unclear.