AI coding agents are exfiltrating your data and no one's talking about it #8767
qo-roj
started this conversation in
Suggestions
Replies: 2 comments
|
Naive question... As a new Omarchy user - if I never use any of its AI features - is my data safe? Or do any Omarchy AI agents silently do any unsolicited AI shenanigans in the background? |
0 replies
|
Educated guess: I'd say yes but where is the point to use Omarchy without any AI features? |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Recently I stumbled over a podcast and learned about Omarchy. Interesting enough but in my eyes there is a crucial part missing.
Every AI coding agent integration discussed here (Claude Code, Codex, Hermes, Grok CLI, Gemini CLI) reads your files, logs, configs, shell history, SSH keys, and API tokens, then ships all of it to cloud LLM providers on every request.
Omarchy ships and integrates these agents. There's no data protection layer between them and the cloud. Nobody in this discussions section has raised it, and no agent harness ships with one either.
I've been building Tidebreak (https://github.com/qo-roj/tidebreak) to fix this — a local redaction proxy (Go, single binary) that:
Agent ──▶️ Tidebreak (localhost:8842) ──▶️ ┌── Cloud API (scrubbed)
└── Local Ollama (full data)
My suggestion: Omarchy should either ship a redaction proxy by default or at minimum document the risk and point users to a solution. Right now people are installing coding agents that have full filesystem access and zero data safeguards, and the community is asking for more agents without questioning what those agents send to third-party servers.
The repo is MIT licensed, contributions welcome. I'd rather this be a community effort than a single-vendor thing.
All reactions