What happens to pasted text.

Plain English, no lawyer theater — this describes what the code actually does, nothing more.

1.What happens to text you paste

When you press Analyze (or Clean), your text is sent to the server, processed in memory, and the result comes back. That text is not stored, not written to a log, and not shared with anyone. When the response is sent it is gone.

One optional action can send part of your text further, and it takes two separate steps. The Dewatermark page has an opt-in checkbox, unticked by default, saying your text may go to an external, third-party LLM provider to have the passages that read as machine-written rephrased. Ticking it on its own sends nothing. After you press Clean, those passages are highlighted in the result, and a Rephrase these passages button appears; only pressing that button, for that document, sends anything — and then only the highlighted passages themselves, at most three of them, one per request each; the rest of your text, and the document as a whole, is never sent. Without that press nothing leaves at all: Analyze and Clean run entirely on this server, in memory.

What the provider receives is the passage text and nothing else — no identifier, no account, nothing tying it to you. What it does with it is governed by that provider's own terms, not by this page, which is why the checkbox is off until you turn it on and why consenting is not the same act as asking. If this deployment has no provider configured, the checkbox is disabled and the option does not exist here at all.

2.What is stored, and when

One thing is stored, and only if you choose to click “AI — I'm sure” or “Human — I'm sure” under the results. That click saves a single line containing:

• the text you pasted,
• the label you picked (ai or human),
• a timestamp,
• the detector scores for that text, the software version that computed them, and a checksum of the text.

Nothing else. No IP address, no browser or device information, no account, no cookie, no fingerprint — the app has none of those. If you never click a label button, this app stores nothing about your visit; the hosting layer underneath it still keeps ordinary short-lived access logs, which is section 7.

The click is the consent: it is a deliberate, optional action, and this page is what makes it an informed one (GDPR Art. 6(1)(a)).

3.What it is used for

Only one purpose: measuring and improving the AI-authorship, disclosure and watermark detectors in this project. The stored text and label are never sold, never shared with third parties, and never used for advertising or profiling.

Labelled submissions may be copied into the project's research dataset on the maintainer's own machine to evaluate detector accuracy. That dataset is private — it is not committed to the repository and not published.

4.How long it is kept

On the server, a labelled submission is kept until it is pulled into the research dataset, or at most 12 months, whichever comes first. Research copies are kept for as long as they are useful for evaluating the detectors, and are deleted when they are not.

5.Your rights

Under the GDPR you can ask for access to your data, correction, erasure, restriction of processing, objection to processing, and portability, and you can complain to your national data protection supervisory authority.

To exercise any of them — including simply having a submission deleted — email coder@unsuckai.com and quote enough of the submitted text to identify the entry. Please include no more personal data than is needed to find it. Erasure covers both the file on the server and any research copies.

6.Before you paste — please read

Do not paste text that contains other people's personal data. You are responsible for what you submit. If you submit someone else's information you may have no right to do so, and the safest option is simply not to paste it.

7.Hosting and server logs

The reverse proxy in front of this app keeps ordinary short-lived access logs, which include IP addresses, for operating and troubleshooting the service. Those logs are infrastructure plumbing, kept separately from label data and never joined to it. There are no analytics, no tracking cookies, and no third-party scripts on this site — every page here loads only its own stylesheet, script and font.

8.Detector scores are not decisions

The scores this site shows are heuristic research output. They are frequently wrong, they are not evidence, and no automated decision with a legal or similarly significant effect on anyone is made here or anywhere else in this project.

9.Who is responsible

The data controller is the operator of this site. This is a personal research proof of concept, not a company; the email address above is the contact channel.

Last updated: 2026-08-24.