The Next Web: Only Z.ai Can Open the Code Z.ai Uploaded

The Next Web reports that code uploaded by Z.ai can only be opened by Z.ai itself. The observation turns on what happens after the upload rather than on the act of publishing, which is what makes the item worth noting on its own.

The point is not that code was made public but whether it can be read, compiled and verified once it is public. For any release marketed as open source, readability is the first step in building trust: if only the publishing party can open the files, outside teams cannot audit, reproduce or build on them.

Information available so far is limited. The contents of the code, the release channel, the timing and the reason a third party cannot open it have not been set out, and there is no indication that Z.ai intends to change how it publishes. What can be confirmed is the single fact the report describes.

Zooming out, open source has long been a blurry term in AI: it can mean usable weights, usable training code, or simply permission for research and limited commercial use. That is why every time a company labels a release as open, the community's first move is to map the boundary of what can actually be obtained, and this report points at exactly that boundary.

For developers and researchers, whether a release is worth their time depends on three things: running it locally, auditing security and licence risk, and fine-tuning on top of it. If any of the three fails, the release falls from infrastructure to a reference implementation.

What to watch next is whether Z.ai responds, whether it republishes readable code or adds licence terms, and what the developer community says in public. If the report's account holds, the approach is also likely to reopen the debate about where the line sits between open weights and open code.