Privacy policy

Last updated: 7 November 2026.

reright is a service of interpt, Av. Nuno Alvares Pereira 52, 2300-532 Tomar, Portugal, VAT number PT253648920. You can reach us at support+reright@interpt.co. In these documents "interpt", "we" and "us" mean that business.

interpt is the controller of the personal data described below.

What we collect

• Account data: your email address, a hash of your password, and email verification and reset tokens.
• Content you submit: the text, context, diff, target and origin of each draft, your edits, the final text, and any reject reason.
• Usage data: a count of rewrites and words per month, your plan, and device tokens with the name you give each device.
• Where you came from: when you create an account, the campaign tags on the link you followed (utm_source, utm_medium, utm_campaign), whether it had a Google Ads click id (only yes or no, never the id) and the name of the site that sent you, such as google.com (never the full address). It is kept with your account and tells us which channels bring people.
• Billing data: subscription status, the Stripe customer and subscription identifiers, and invoice records. Stripe takes the payment and calculates tax. Card details go to Stripe and never reach us.
• Notification data: your browser push subscription if you enable push.
• Feedback: if you send feedback from the app, we get your message by email together with your account email, plan, usage, number of devices, the page you were on and your browser. We use it to read and answer you, and keep it in our mailbox.
• Your writing: text samples from the sources you connect, the AI versions of those samples made for training, the pairs of agent text, rewrite and the text you approved, the list of connected sources with the GitHub access token, and your rewriting settings. For a Gmail reply we also keep, for a short time, the message it answers: that message's text, the sender's name as shown in the mail (not the address), its subject and its message ID. That message was written by someone else; see the retention section for how long it stays. The text of samples, pairs and answered messages, the AI versions and the source tokens and details are encrypted like draft text. The rest, such as the source, kind and dates of each sample, is not.
• Technical logs: IP address, request time, method, path and response status, kept by the server and by Cloudflare, which carries the traffic for the app and the marketing site, to run and secure the service. The server deletes its access logs after 7 days, and backups drop them within 48 hours after that.
• Try box on the site: the text you paste goes to Modal to be rewritten, with the same masking and swap as any rewrite, and comes back to your browser. reright does not store it or write it to logs. To enforce the limits we keep keyed hashes of the box's cookie id, your IP address and your network for 24 hours, and a count of tries and words for each day.

The marketing site at reright.it sets no cookies and runs no analytics. When you arrive from another site or from a tagged link, its links to the app carry the tags and the name of that site, so the app can record them if you sign up; nothing is stored in your browser. It loads no third-party scripts until you click into the try box on the home page. Then it loads Cloudflare Turnstile, which checks that you are a person, and app.reright.it sets a cookie holding a random id, which counts your tries for 2 days. The app sets a session cookie that is needed to keep you logged in. Your theme choice is kept in your browser storage and is not sent to us.

Why we use it and on what basis

• To provide the service you asked for, which is performance of a contract: accounts, drafts, approvals, notifications and quotas.
• To rewrite your text so it sounds like you, and to learn from your writing and edits for you alone, which is performance of a contract.
• To use the messages your Gmail replies answer as context when we make training examples for your model, which is our and your legitimate interest. The people who wrote those messages are not our users, so we keep their text only briefly, swap names and other details before it reaches Modal's models, and never train on it by itself.
• To take payment and keep tax records, which is performance of a contract and a legal obligation.
• To keep the service secure and stop abuse, which is our legitimate interest.
• To know which channels bring new accounts, from the sign-up source described above, which is our legitimate interest.
• To send you emails about your account, such as verification, reset and billing notices. We send no marketing email.

Where your text is kept

Traffic between you and the service uses TLS. Cloudflare ends the TLS connection for the app, so your text passes through Cloudflare in transit. The application and its database run on Scaleway servers in Amsterdam. Draft text, context, diffs, targets, origins, final text and reject reasons are stored encrypted with AES-256-GCM. Each user has a separate data key. Those keys are stored wrapped by a master key that lives on the server outside the database, so the text fields in a copy of the database or of a backup cannot be read on their own. Other data is not encrypted in the database: your email address, your plan, the kind and status of each draft, timestamps, notification settings and push endpoints, billing status, your rewriting settings, the source and kind of each writing sample and pair, and the status and results of training runs.

This is not end-to-end encryption. The server decrypts your text to show it to you and to check approvals, and interpt operates the server, so someone with full access to it could read your text. We limit that access to what running the service requires.

Approval checks use a keyed hash of the text, not the text itself.

Rewriting

Rewriting runs on Modal (Modal Labs, Inc.) in the United States. The models are open models that we run there ourselves. We do not send your text to another company's AI service.

Before text leaves reright, the server replaces code, links, email addresses and @mentions with placeholders, and puts them back when the rewrite returns. On Modal, a detection model then looks for names, usernames, phone numbers, bank and card numbers, tax and ID numbers, street addresses, dates of birth, IP addresses and similar details, and swaps each one for a made-up value of the same kind before the rewriting model sees the text. The same real value always gets the same made-up value in your account, and the real values are put back before the rewrite comes back to reright. The swap is worked out from a secret key each time and is never stored. Detection can miss a detail, so do not treat the swap as a guarantee.

Modal gets your text only for the job at hand, and we never write it to logs. Modal itself keeps the requests we send it and the answers it returns for up to 7 days, as part of how it runs our code, and then deletes them. Those requests hold your text as it was before the swap: for a rewrite, the text to rewrite, and for training, the samples and, for Gmail replies, the messages they answer. Once a training request arrives, the job stores it encrypted, swaps the personal details, works on the swapped copies in Modal's storage and deletes them when it ends, and a daily sweep removes anything left behind by a job that crashed. Your model, a small add-on to the shared base model, is kept on Modal under a random key that is not your user id or email address. Documents you upload in PDF, Word, OpenDocument or RTF format go to Modal as they are, so that it can read their text, and are not kept there beyond those 7 days. The masking and the swap apply to the text read from them.

Your text trains only your own model. The shared base model is trained on public text, not on anything our users send. A rewrite that went back to an agent without anyone reviewing it is never used for training.

You choose the sources reright learns from:

• Gmail: you export the mail you sent with Google Takeout and upload the file. reright never connects to your Gmail account. The server reads the messages you sent, and the file stays on its disk, unencrypted, only until it has been read. Then it is deleted. From each reply we keep your own words and, until the next training run, the message you answered. That message goes to Modal with the training request, after the same masking and swap, and may be used there to write a reply the way an assistant would, which your model then learns to turn into your real reply.
• GitHub: through reright's GitHub App, with read-only access to the accounts and repositories you install it on. reright reads the commit messages, pull requests and comments you wrote, and reads them again every week while the app is connected. The access token is stored encrypted, and disconnecting revokes and deletes it. Disconnecting keeps the samples already taken; you delete those on the Your writing page.
• Google Docs: you pick documents in the Google Picker, which gives reright access only to the files you pick (the drive.file scope). reright exports them as plain text and does not keep the Google access token.
• Files you upload: kept on the server's disk, unencrypted, only until their text is read, and then deleted. An upload that never finishes is deleted within about a day.

GitHub and Google are not our subprocessors. When you connect them, you use your own account with them, and each handles that under its own terms and privacy policy.

reright's use and transfer of information received from Google APIs to any other app will adhere to the Google API Services User Data Policy, including the Limited Use requirements. Text from your Google Docs is used only to rewrite your own text and to train and run your own model, never the shared base model or anyone else's model. Nobody at interpt reads it unless you ask us to, it is needed to investigate abuse or a security problem, or the law requires it.

Retention

We delete message text from the live database 30 days after the decision. Text is kept 30 days on every plan, Free included. A draft that nobody decides expires after the same period, counted from when it was submitted, and its text is deleted then. Counts and billing records stay. The deletion job runs every hour and once each time the server starts.

Text is deleted from the live database after 30 days, and from every backup within 48 hours after that. Backups hold the database as it is, with the text fields encrypted as described above. The database is streamed to Scaleway Object Storage in Amsterdam, which keeps 24 hours of history and removes anything older than 2 days. The server keeps no nightly copy. Point-in-time recovery therefore only goes back about a day. A database restored from a backup is purged again when the server starts, before it answers any request.

When you delete your account, your data is removed from the live database right away and from backups within 48 hours. Any paid plan is cancelled first. We keep a list of one-way hashes of the email addresses of deleted accounts, so that restoring a backup does not bring a deleted account back. The list holds nothing else.

Your writing is kept until you delete it. The messages your Gmail replies answer are the exception: each is deleted once a training run has used it, and in any case 30 days after the import, and it leaves the backups within 2 days after that. On the Your writing page you can delete one source's samples or all your writing. Deleting all your writing or your account deletes the samples, pairs and source tokens from the live database right away. Within about 15 minutes reright asks Modal to delete your model, and keeps asking until Modal confirms. Copies of recent requests that Modal keeps on its side are gone within 7 days. We keep a record of each of these requests, holding only a random key and the internal number of your account. Deleted writing leaves the backups within 2 days. The rewrite stored on a draft follows the draft's retention. Pairs do not: they keep a copy of the agent text, the rewrite and the approved text until your writing is deleted.

Who receives data

We use the providers on the subprocessor list. We do not sell personal data. We do not use your text to train models, except your own model.

If you turn on push notifications, the notification goes through your browser vendor's push service. By default it holds the kind of draft and its target. If you turn on previews, it also holds the first line of the text.

The reright command and its hook program are open source and are downloaded from GitHub releases (github.com/interpt-co/reright-cli), not from our servers. GitHub sees the IP address of the download and handles it under its own privacy policy. We do not receive it.

Transfers outside the EEA

Some providers, including Cloudflare, Stripe and Modal, may process data in the United States or other countries. Where they do, we rely on the European Commission's standard contractual clauses or an adequacy decision.

Your rights

Under the GDPR you can ask for access to your data, correction, deletion, restriction, a portable copy, and you can object to processing based on legitimate interest. You can delete your account and all its data yourself in Settings. For anything else write to support+reright@interpt.co. You can also complain to your national data protection authority.

Security incidents

If a breach affects your personal data in a way that creates risk for you, we will tell you and the authorities as the law requires.

Children

The service is for developers and is not aimed at children.

Changes

If we change this policy in a way that matters, we will email account holders before it applies.