Why your drafts should never be stored

A rewriter reads everything you’re about to say. Think about what passes through that pipe in a normal week: fragments of work documents, half-formed product ideas, personal questions, code you haven’t shipped, names of clients and colleagues. Drafts are the rawest text you produce. They are less filtered than anything you actually send.

Most tools in this category treat that stream as an asset. The text goes to their server, gets kept, and you take the privacy policy on faith. Even with the best intentions, stored text is a liability: it can be logged, it can be breached, it can be subpoenaed, and a retention policy can change after you’ve built the habit.

The posture: never stored, never trained on

RefineIQ takes a different route. The rewrite runs on a secure RefineIQ backend, and the rule is strict: your text is never stored, never logged, and never used to train any model. It exists only for the millisecond it takes to rewrite, then it is dropped. Nothing is retained to log, breach, or subpoena.

Running the rewrite on a backend is what lets the modes be good. A server-side model does the heavy lifting so “Sound Human” can make a draft read like a person wrote it, and “Clean up” can fix the grammar and cut the filler without changing your voice. A quality gate strips any preamble the model adds and falls back to your original text if the rewrite adds nothing.

Voice stays on your device

Voice Draft goes further than the text path. Dictation is transcribed on-device by a small multilingual speech model running in your browser. Audio is never stored and never uploaded. Only the transcribed text is sent to the backend to be refined. RefineIQ deliberately avoids the browser’s built-in speech recognition because Chrome implements it with a server-side service.

How to evaluate any tool that reads your drafts

Three questions cut through most privacy theater:

1. Does it store your text? If the answer is yes, your drafts are sitting on someone else’s disk.
2. Does it log your text? Request logs are text too. A tool that logs every rewrite has a copy of everything you wrote.
3. Does it train on your text? “We may use your data to improve our models” means your drafts become training data you can’t pull back.

RefineIQ’s answers are no, no, and no. That’s what “private by default” should mean. The full accounting of what is (and isn’t) handled lives in the privacy policy.