Realtime AI News
Zhipu's ZCode Upload Mechanism Exposed, With User Codebases Reportedly Packaged and Sent
A September 20 report says the upload mechanism inside Zhipu's ZCode coding tool has been exposed, with user codebases packaged and uploaded without clear notice. The claim comes from third-party analysis, and no public explanation from Zhipu has appeared so far.
On September 20, a Chinese-language technical blog aggregated through Google News said the upload mechanism inside Zhipu's ZCode coding tool had been exposed — user codebases packaged and uploaded without notice.
What makes this sensitive is the payload. A codebase is not just text: it usually carries private business logic, internal endpoint addresses, keys and configuration fragments. Any unexpected upload can move a company's most sensitive material across a boundary it never agreed to cross.
Technically, the question is not whether uploads happen but where the boundaries sit: how much is sent, meaning an entire repository or selected files; what triggers it, on by default or confirmed by the user; and how the data is used and how long it is retained.
Coding assistants and agents generally need to process context in the cloud, so uploading is not inherently out of bounds. The industry norm is to write collection scope into a privacy policy and to surface explicit prompts and toggles at first use, or when repository-level indexing is enabled.
Only headline-level description is visible today, with no official statement and no independent verification, so it is hard to judge whether this is a design flaw, inadequate notification, or a mismatch between documentation and actual behavior.
Watch whether Zhipu explains upload scope and data use, and whether it offers opt-outs or local processing. For enterprises bringing coding agents into internal repositories, the dispute is a reminder to confirm data flows before deciding how much access to grant.
Why it matters
If upload scope and triggers cannot be stated clearly, enterprises will hold back from connecting internal repositories to coding agents, making data-flow transparency a gate for entering the enterprise market.
Nearby Updates
All09/20, 18:39
Zhipu discloses GLM-5.3, saying the model is starting to optimize the inference system that runs it
A report published by OSCHINA says Zhipu has disclosed that GLM-5.3 is beginning to optimize the inference system that carries it, an early sign of recursive self-improvement, or RSI. The claim pushes the idea of a model improving its own execution stack from theory toward a concrete product statement.
09/20, 18:50
Another Chinese MoE flagship lands third in the open-source ranks, with pricing aimed at DeepSeek-V4-Pro
Zhidongxi reports that another Chinese MoE flagship model has arrived, placing third among open-source models worldwide while pricing itself against DeepSeek-V4-Pro. The launch raises the stakes on both the leaderboard and the price competition among Chinese open-weight releases.
09/20, 15:56
Bilibili launches AI Infinite Arena leaderboard with 100 models competing
Chinese tech coverage reports that Bilibili has launched an evaluation leaderboard called AI Infinite Arena, describing it as one competition in which one hundred models from around the world take part. It marks the video platform's first move into ranking models, giving Chinese developers and creators a community-owned reference point beyond vendor self-reports and overseas arenas.
09/20, 14:48
Former OpenAI Researcher Releases Jev, a Model for Fast, Structured Software Decisions
According to OSCHINA, a former OpenAI researcher has released a model called Jev that aims to help software make fast, structured decisions. It is another attempt to pull decision-making out of general-purpose chat models, though public details currently stop at the announcement itself.