Last updated: 14 August 2026
The rest of this page explains the detail, including the cases where more is transmitted than you might expect.
When you translate a page, NativeRead extracts the visible text of that page and sends it to our servers for translation. This is the literal text as it appears — not a summary or a sample.
Extraction is not limited to article prose. It covers paragraphs, headings, list items, table cells, form labels, captions, and general text-bearing containers. In practice this means that if text is visible on the page, it is likely included.
Please read this before translating a sensitive page.
NativeRead has no list of excluded websites. If you translate a page on a banking, healthcare, workplace, or other private site, the visible text of that page is transmitted like any other page.
Text sitting inside rich-text editors — a partly written email in a webmail composer, a chat message box, a note-taking app — is treated as page text and may be transmitted. Text you type into ordinary form fields, search boxes, and password fields is not read.
The full URL of the page being translated is sent with every translation request, and is used to count how many distinct pages you have translated against your plan limit. URLs can contain search terms, identifiers, and query parameters, so we treat them as personal data.
We do not write page URLs to our database. They are held in our server's working memory, associated with your account or IP address, and cleared at midnight UTC each day. A shortened form of the URL (roughly the first 50 characters) also appears in our operational server logs.
If you create an account, we collect your email address and display name. If you sign in with Google, we receive your name and email address from Google. Authentication is handled by Firebase Authentication.
NativeRead does not create a device identifier, install identifier, or advertising identifier.
Payments are processed entirely by Stripe. We never receive or store card numbers or bank details. We store only your Stripe customer ID so we can link your subscription to your account.
This is the part most people want to be precise about, so it has its own section.
| Trigger | What is sent |
|---|---|
| You click the translate button, or use the popup | The visible text of that page, plus its URL. |
| You have enabled auto-translate | Every page you subsequently visit is translated automatically, shortly after it loads, with no further action from you. |
| A page you already translated loads more content | Newly added content on that page — from infinite scroll, new chat messages, or in-page navigation — is translated automatically while the page remains open. |
Auto-translate is off by default and can only be turned on by you. Be aware that when it is on it applies to every site — there is no per-site control and no allowlist. If you enable it, pages behind a login will be translated automatically along with everything else.
NativeRead reads only the top-level page. It does not read content inside embedded third-party frames.
input, textarea, or select elements, including password fields.The language of the page decides where your text is processed. For most languages the translation runs on our own infrastructure — a self-hosted model on servers we control — and your text is not shared with any third-party AI company.
For some source languages our own model translates poorly, and those are sent to OpenRouter every time, not only during a fault. The same applies to languages our own model does not support at all. Which languages these are changes as we measure quality; the current list is in our source code. A request our own model fails to complete — an outage, a timeout, maintenance — also goes to OpenRouter, whatever its language.
In every one of those cases OpenRouter routes the request to Cloudflare, which is the only company permitted to run the translation. The models themselves are published by Google and DeepSeek, but neither company receives your text — Cloudflare runs the models on its own infrastructure.
This is enforced in our code rather than left to configuration, in two independent places: every request names the companies permitted to run it, and any model whose publisher is not listed above is refused before a request is sent. Both matter. OpenRouter offers these models through more than a dozen companies and picks one itself unless told otherwise, so without the first restriction your text can reach a company named nowhere on this page.
No Google AI service translates your text. Google is still involved elsewhere and we do not claim otherwise: our backend runs on Google Cloud, our database is Google Cloud SQL, and sign-in uses Firebase. Those are listed under "Other services" below, and your page text passes through that infrastructure on its way to and from us.
This routing is automatic and happens without prompting you.
Cloudflare and OpenRouter each handle data under their own privacy terms and retention practices, which we do not control and make no claims about. If retention matters to you, review their terms before translating sensitive material.
| Data | Retained |
|---|---|
| Page text you translate | Not stored by us. Held only for the duration of the request. |
| Translated results | Not stored by us. Cached only in the page's own memory while that tab is open. |
| Page URLs | In server memory until midnight UTC. Partial URLs in server logs, subject to our hosting provider's log retention. |
| Account information | Until your account is deleted. |
| Usage and billing records | Kept while your account is active, and afterwards where needed for accounting, support, and legal obligations. |
You can do the following directly in the extension:
For anything else — including access to the full set of data we hold about you, correcting your email address, exporting your data, or deleting your account — email liang@trendable.no. We will respond within 30 days.
We handle these requests manually at present rather than through a self-service control. We would rather tell you that plainly than imply a button exists that does not.
Two things to note about deletion. Records that are no longer linked to you — anonymised usage and token-accounting rows — may be retained after your account is removed. Your customer and billing records held by Stripe are subject to Stripe's own retention policy and their legal obligations.
All data is transmitted over HTTPS/TLS. Authentication uses signed tokens. API keys and service credentials are held server-side and are never included in the extension.
No system is perfectly secure, and we do not claim otherwise. If you believe you have found a vulnerability, please email liang@trendable.no.
NativeRead requests these Chrome permissions:
NativeRead's use and transfer of information received from Google APIs adheres to the Chrome Web Store User Data Policy, including the Limited Use requirements.
NativeRead is not intended for children under 13, and we do not knowingly collect personal information from them.
Our servers are hosted in the United States, and the AI providers listed in section 4 operate globally. If you use NativeRead from outside the United States, your data is transferred and processed there.
We may update this policy. If a change materially affects how your data is handled, we will update the date at the top of this page and notify users in the extension.
For privacy questions or data requests: liang@trendable.no