Trust & privacy
Exactly what happens to your writing — in plain language, with the technical detail a tap away.
What happens to a chapter
You paste a chapter and choose which partners should read it.
Critique Partner’s own system builds your critique and keeps your work in your private account.
Our system works through leading AI models to draft the words. They never train on your writing.
Your critique returns to your browser the same encrypted way — and your writing is never sold, shared, or used to train AI.
Three steps, one closed loop — your chapter goes only where it must to earn you feedback, and nowhere else.
Our principles
Not for us, and not for the AI providers we use.
Chapters go to established AI providers — such as Anthropic and OpenAI — through their commercial APIs, not consumer chat apps. Under those terms, submitted content is not used to train their models by default, and we add no training pipeline of our own.
Unreadable as it travels, and encrypted on the disks where it lives.
In transit, all traffic uses HTTPS / TLS and session and CSRF
cookies are marked Secure. At rest, your data sits on storage that
is encrypted by default through our hosting provider, Railway,
which is SOC 2 Type II certified.
We log what a critique cost — never what it said.
Every AI call is logged for metering, but the log keeps only token counts, cost, latency, and success/failure — never the prompt or your writing. Your manuscript lives only in your project.
Visible to you alone — never sold, never handed to advertisers.
Your work is scoped to your authenticated account; other users and the public can’t reach it. Our own team accesses it only to operate the service, help you when you ask, or meet a legal obligation. The only outside services that ever touch your text are the trusted AI models that generate your feedback.
Our own critique system, built on leading models — not a consumer chatbot.
The way we shape and focus your feedback is proprietary. The underlying models come from established providers — such as Anthropic and OpenAI — via their developer APIs under commercial terms, so we keep the freedom to use the best model for each task.
Read more
This describes how Critique Partner works today — we won’t claim a protection until it’s real.