Your data

Your words get rewritten. They don't get kept.

Runs on our own model. Your text never goes to OpenAI, Anthropic, or any third party.

RefineIQ's rewrite modes run on our own backend — not a wrapper around someone else's API. The trade we made for that: your text is never stored, never logged, and never used to train a model. It exists for the moment it takes to rewrite, then it is gone.

You select text

In Slack, Reddit, GitHub, a support box, or any text field, the Refine pill appears where you work.

The rewrite runs, then the text is gone

Your draft goes to the backend over HTTPS, gets rewritten, and comes back. It is never stored, never logged, never used to train a model.

You review and replace

A plain-text rewrite lands in place. Esc restores the original. Voice Draft is transcribed on your device, so audio never leaves it.

What stays on your device

Voice Draft is transcribed locally. A multilingual speech model runs in your browser. Audio is held in memory only while it is transcribed, then dropped. No recordings, no files, no cloud speech service.
Your settings and per-site rules. Preferences live in the browser, not on an account profile we read.

What goes to the backend, and what we do with it

Only the text you choose to refine. It travels over HTTPS, gets rewritten, and comes back. Nothing else rides along.
Never stored. The request is not written to a log or a database. There is nothing to breach after the fact.
Never trained on. Your drafts do not become training data. That is a rule, not a setting you have to find.

What we ask Chrome for, and why

Permissions are justified host-by-host in the privacy policy. Short version: storage for your settings, optional microphone for Voice Draft, and site access so the Refine pill can appear where you type.

Go deeper

Never stored. Never trained on. ✎

Add to Chrome