These terms explain what the Blood Swipe Classifier stores when you submit an image for analysis, how long it is kept, when we will email you, and the choices available to you. Creating an account means agreeing to them.
For each classification, depending on the mode you choose (see Section 3), the following may be stored in our database:
Those quality checks are the one thing recorded even in private mode. They describe the analysis rather than the picture, and they attach to the result row rather than to any image, so no image is retained by them. We say so here rather than leave it to be discovered.
If you use the case likelihood-ratio tool, we store something different: the small crops you cut out of your photographs, the label you gave each one (proposition A, proposition B, or questioned), the names and descriptions of your two propositions, any object-reference photo you add, and the report the tool generates. The source photographs themselves are never stored. They stay on your computer; only the crops you choose to take from them are uploaded.
Separately, your account record stores your email address, full name, and organisation. Account data is kept for as long as your account exists and is removed when the account is closed.
Your account record also stores whether you receive occasional email updates about the tool, when that was set, and where it was set (from these terms at registration, from your account, or from an unsubscribe link). Four things about that list are true by design:
| What you used | Images stored? | Retention of images | Credit cost |
|---|---|---|---|
| Classifier - Standard (incl. Detailed) | Yes | Erased 30 days after submission | Base (1 / 2 / 3) |
| Classifier - Private | No - never stored | Not applicable | Double (2 / 4 / 6) |
| Case likelihood-ratio tool | Crops only - never the source photograph | Kept indefinitely | Free |
In every classifier mode, the result and basic metadata stay in your history. In standard mode only the images are erased after 30 days; the result row remains, so your history stays complete - you simply can no longer view or download the images for that entry.
Research mode was withdrawn on 12 August 2026; submissions made under it before that date remain subject to the terms then in force (image and ground truth kept 3 years, or for as long as the dataset is used where the submission was incorporated into one).
Your uploaded image and the generated images are stored so you can review and download them. They are automatically and permanently erased 30 days after the date of submission. After that the history entry still shows your result and metadata, but the images are no longer available. If you need the images for longer, download them within the 30-day window or use a different mode.
No image is ever written to our database. The image is processed in memory only; the result and metadata are recorded, but there is nothing to retain or delete afterward. Private mode costs double credits because it forgoes the storage that would otherwise let you revisit the images. Choose this mode for sensitive or confidential casework.
This tool is experimental and is not validated for casework. It is offered on the condition, which you confirm with a tick box before each comparison, that you understand that and that we may reuse the case.
What we keep is the crops and their labels - each stain image, whether you filed it under proposition A, proposition B or "questioned", the names and descriptions of your two propositions, any object-reference photo, and the generated report. Your source photographs are never uploaded and never stored.
This material is retained indefinitely. That is deliberate rather than a default: a set of stains labelled by an examiner as coming from the same source or from different sources is a reference set whose value is that it lasts, and it is the kind of data the classifier cannot get any other way. It may be used, in anonymised form, to extend and validate the models. It is not sold and not used for advertising.
You can download any case at any time as a single archive (the crops, their labels and the report), from the "download this case" link in the builder. You can also ask us to delete a case - see Section 5.
Research mode was withdrawn on 12 August 2026; submissions made under it before that date remain subject to the terms then in force. Those submissions - the image, the generated images and the ground-truth entries - are retained for 3 years from submission, and where a submission was incorporated into a training or validation dataset it is retained for as long as that dataset is used, potentially indefinitely, including in anonymised or aggregated form.
Data you contribute - historically through research mode, and today through the case likelihood-ratio tool - is only worth keeping if it is accurate. To protect the integrity of that reference set:
If you are unsure whether you can label a sample honestly, say so in the proposition description rather than guessing.
To make any of these requests, contact us at the address in Section 7. If you are in the EU/EEA, these choices support your rights under the GDPR, including access, rectification, and erasure. Forensic swipe images may in some contexts constitute personal or sensitive data; you are responsible for ensuring you have a basis to upload any image, and for using private mode where confidentiality is required.
Stored images and results are held in our application database. Access is limited to your own account and to administrators for support and operational purposes. We do not sell your data or use it for advertising.
Questions about these terms or requests regarding your data: daniel@attinger-consulting.com
We may update these terms as the service evolves. Material changes will be reflected in the "Last updated" date above and, where appropriate, communicated in the app.