Grok Build Exposes Code Upload Risks
Substantial_Step_351 · reddit · 2026-07-15
This discussion focuses on a security/privacy issue in Grok Build. According to an article by cereblab, the CLI uploads the entire git repository, including full history, to xAI's storage buckets.
Key Points from the Article
- In a roughly 12GB test repository, model-related traffic was only about 192KB, but the storage channel transmitted 5.1GB.
- It even uploaded a file explicitly instructed not to be opened.
- More concerningly, the fix wasn't a client update; the next day, the server-side simply flipped a switch: disablecodebaseupload true.
- This indicates the capability was remotely controllable all along, rendering local "privacy switches" unreliable.
Core Discussion Points
The post emphasizes that while people focus on prompt injection or model behavior, the real danger lies in the harness / tool layer:
- It has network access
- It can access your entire repository
- It might feature auto-update channels
- The actual risk might be higher than random npm packages
The author concludes by asking if anyone is performing egress filtering for coding agents or actually inspecting traffic payloads.
More from coding & agent
- AIE Code Summit 2026 opens applications: 800 seats, keynotes streamed free — trq212 · 2026-10-27
- Decade-long cloud architect: enterprise AI architecture is far more than bolting on an LLM — kashifmanzoor · 2026-10-27
- Analysis of thousands of coding sessions: 40-89% of files agents read are never used — rseroter · 2026-10-27
- TRL adds Score Centering, a simple fix for RLHF training instability — lvwerra · 2026-10-27
- Precompute golden answers and match at runtime for exact wording control — colinmcnamara · 2026-10-27
- Dev argues there's still alpha in working with your coding agent interactively — alex_frantic · 2026-10-27