Privacy

The short version

ProtoGloss does not train on your documents. Your text is processed to translate it, and it passes to a configured model backend to do that. Where that matters most, and where the honest answer depends on your deployment, is set out below rather than glossed over.

What we collect

If you contact us, we hold what you send: your name, your email address, and whatever you write in the message. We use it to reply to you and to run the evaluation you asked for. Nothing more.

If you have workspace access, we hold the documents you upload, the glossary and character material you create, the translations produced, and the operational records needed to run and bill the work.

What happens to an uploaded document

Uploads land in a quarantine area first. They are validated by size and file signature before being promoted for parsing. Parsing itself runs in a restricted workload that holds no model-provider credentials and has limited network egress, so extraction happens before, and separately from, any model call.

We do not scan uploads for malware. The interface exists but currently ships as a permissive stub, and we would rather tell you that than imply a protection you do not have.

The model backend, and why this answer is conditional

Translating your text means sending it to a configured model backend. That backend's own retention and training policy governs that step, and the backend is a deployment-level choice rather than a fixed property of the product.

This is the question every publisher should ask, and the only honest general answer is that it depends on the deployment you are on. If it matters to you, and for a manuscript under contract it should, ask us in writing and we will tell you exactly which backend your workspace uses and what its terms say.

Retention

Command line usage defaults to thirty day retention, with deletion that previews what it will remove before it applies. For hosted workspaces, retention is agreed with you rather than imposed silently. You can ask us to delete your material and we will.

What we do not do

We do not sell your data. We do not use your manuscripts to train ProtoGloss. We do not publish or share customer material, and we do not name customers without written permission.

Analytics and the website

This site is deliberately light. We do not run advertising trackers on it.

Contact

Questions about any of this go to [email protected], and we would rather have the conversation than have you assume.

Status of this page

ProtoGloss is an early-stage product operated from Canada. This page describes current practice in plain language and will be replaced by a fuller policy as the product and the company formalise. If you need contractual assurances now, ask and we will put them in writing.