ZCode's silent git uploads show why agents need network audits

Two independent investigations found that ZCode, a GLM-based coding agent, has been quietly uploading users' git history and full workspace snapshots to remote servers. The findings prompted a warning that users should check what their coding agent sends over the network before trusting it with a private repository.

ZCode, a GLM-based coding agent, has been quietly uploading users' git history and full workspace snapshots to remote servers, two independent investigations found. Check what your coding agent sends over the network before trusting it with a private repo. Read: ZCode, a GLM-based coding agent, has been quietly uploading users' git history and full workspace snapshots to remote servers, two independent investigations found. Check what your coding agent sends over the network before trusting it with a private repo. Read: MiniMax released v0.4.12 of its Code CLI as open source under the MIT license, publishing the full agent harness for developers to inspect or extend. Read: SemiAnalysis describes Engram, a model architecture that extends token embeddings with learned lookups addressable by token ID, letting rows prefetch from host DRAM instead of living in HBM. Read: Claude Code 2.1.277 falls back to reading a project's AGENTS.md file when no CLAUDE.md exists, aligning it with the cross-tool AGENTS.md convention for agent instructions. Read: CNN reports a US military unit nearly escalated based on an AI system's hallucinated intelligence about a Chinese ship, an incident surfaced through internal review. Read: TypeSafe AI's Jev, a probabilistic decision model for agent tool-routing and guardrails, became the fastest-adopted model in Vercel AI Gateway history within 24 hours of launch. Read: Epoch AI tracked AI acknowledgment across arXiv math preprints rising from 4% in April to 25% in August 2026, holding even for authors with pre-2023 publication histories.