Zhipu, touted as the "pioneer in global large-scale AI models," has recently been embroiled in a trust crisis stemming from its AI programming tool, ZCode. Developers uncovered that ZCode was covertly packaging project data and attempting to upload it to the cloud. In response, Zhipu attributed the issue to an inadvertent activation of the "codebase indexing feature" and issued an apology for the incident. Nevertheless, its assertion of "data destruction" failed to gain widespread acceptance.

Subsequently, Chengming Technology sent a letter, alleging that ZCode had uploaded its entire archive files, which contained core business data. The company inquired whether the data had been completely erased and whether it had been leaked externally. Zhipu has yet to respond to these allegations.

Industry experts opine that the incident is more likely a result of a loss of control over the implementation and security filtering of the codebase indexing feature, rather than a deliberate act of data theft by Zhipu. However, issues such as the overly broad scope of data uploads and the absence of sensitive information filtering have breached the sanctity of corporate core data assets.

Zhipu's commercialization strategy primarily hinges on B-end clients, and this incident has sown seeds of doubt among developers regarding the company's trustworthiness. It may also catalyze a shift in the AI programming sector, steering the competition away from functionality and speed towards a greater emphasis on security and trust.

Currently, Zhipu has announced that its MaaS open platform will incorporate a "no data retention" feature. To regain trust, Zhipu must surmount three key challenges: clearly delineating the boundaries of data processing, ensuring that data uploads cease when users disable relevant features, and providing irrefutable evidence that the uploaded data has been thoroughly deleted.