Privacy policy
Last updated: 5 September 2026
Short version: we keep your material for 30 days so you can come back for it, then delete it. One exception is described in section 4, and we would rather you read it than discover it.
1. Who is responsible
unmarkclaude.io. Questions about your data, including deletion: hello@unmarkclaude.io. Payment data is handled by lava.top, not by us; see section 5.
2. What we collect
- The material you submit: the text, code or file you want cleaned, and the result we produce from it. Anything you paste into the mark tester counts as material too: 30 days later all that is left is the outcome of that check and, if you named it, the tool you ran the sample through, with no text and no IP attached.
- Your email address, but only if you pay: it is where the link to your result is sent. Before payment we ask for nothing.
- Basic request data: IP address and timestamps, used to enforce the free preview limit and to keep the service from being abused.
- Payment records: what was paid, when, and for which job. Card details never reach us; they go to the payment provider.
- Comments you post: the name and text you type are meant for publication and appear on the page if the comment passes review. Alongside them we record the IP address and browser signature of the submission; those exist only to keep robots out, are never published, and are erased as soon as the comment is reviewed, and within 14 days of posting in any case. A published comment stays until you ask us to remove it at hello@unmarkclaude.io.
There are no accounts and no passwords on this service, so there is nothing to leak.
3. How long we keep it
Your material and your result are kept for 30 days and then deleted. That window exists so the link in your email still works if you close the tab. After it passes, the result page stops working, because there is nothing behind it any more.
Payment records are kept longer, because tax and dispute rules require it. That record is the transaction, not your text.
4. The one exception: the anonymised archive
We keep an anonymised before-and-after pair from each job beyond the 30 days. It is stripped of your email, your name, your IP and your payment record, and it is not linked back to you.
The reason is the promise this whole service rests on. Nobody can verify today whether a mark survived processing, because Anthropic has no public detector. On the day it opens one, we intend to run our archive through it and publish what we find, including the parts that make us look bad. Without the archive that promise is words.
If you would rather your material was not kept for that, say so at hello@unmarkclaude.io with your job reference and we will delete the pair. This costs you nothing and does not affect your result.
Ask within the same 30 days, while your job is still on our side. At the end of that window the job number is cut off the pair for good, and that is deliberate: from then on nobody, us included, can point at a pair and say whose it was, so there is nothing left to look up and nothing left to ask us to remove. What stays in the archive is a before-and-after text with nothing of you attached to it.
5. Who else touches your material
Your material never goes to Anthropic, OpenAI or Google. That is not only a privacy position, it is a technical requirement: sending your text to a large commercial model would hand it to that provider and stamp their own mark on it in place of the one you came here about. Processing runs on open-weight models only.
| Who | What for | What they see |
|---|---|---|
| OpenRouter (routes to open-weight model providers) | Running the open-weight model that rewrites text | The text segments being processed |
| lava.top | Taking payment | Your payment details and email. Not your material. |
| Resend | Sending you the link to your result | Your email address. Not your material. |
Files whose processing is purely mechanical (stripping C2PA provenance metadata from images) never leave our own server at all.
6. Cookies and analytics
Since 26 August 2026 this site runs two visitor counters: Google Analytics 4 and Yandex Metrica. They exist to tell us how many people arrive and how far they get. Four things are counted, the same four for everyone: the page view, then material pasted, result shown, checkout opened.
What they receive: the address of the page you are on, the page you came from, your IP address, your browser and device, and cookies of their own that recognise a returning visit. Yandex Metrica also records the session itself: clicks, scrolling, mouse movement and the pages as they looked on your screen.
What they do not receive is your material. The field you paste into and the block your result appears in are marked with the counter's own do-not-record attributes, so their contents stay out of the session recording. The counted steps carry no part of it either: a size in characters or bytes, whether it was text or a file, and the job number. Any content blocker stops both counters, and nothing on this site depends on them working.
There are no advertising cookies here and no cross-site advertising audiences. If that ever changes, this section changes with it on the same day.
7. Your rights
You can ask what we hold about you, ask for it to be deleted, or ask for a copy, by writing to hello@unmarkclaude.io. Since we hold almost nothing and hold it briefly, these requests are usually answered the same day.
8. Security
Traffic runs over HTTPS. Material is stored on our own server, not in a public bucket. Access is limited to the operator. We will not pretend to certifications we do not have.
9. Changes
If this policy changes, the date at the top changes with it. Material already submitted stays under the policy that was in force when you submitted it.