We delete the file you upload 30 days after it arrives, automatically. What we made from it — your report, your trained model and its evidence receipt — is yours, and we keep it. You can delete any table at any time, and your whole account, uploaded files and run results on request. For Free and Pro, uploading data permits us to retain de-identified copies, including output from our initial de-identification process, for 18 months or more for model improvement, training and quality assurance. These copies remain sensitive and access-limited; they have not been approved against full de-identification criteria. As those criteria are developed, we will re-check earlier copies and delete those that fail.
While a run is happening
Your dataset goes into a container created for that run and nothing else. That container has no network access — no internet, no other machines, not even name resolution. The only thing that can reach it is you, signed in, through this product. Nobody else's run shares it.
When the run finishes we copy out the report, model, and predictions, and then destroy the container. If you chose to keep the session open, it is destroyed after an hour of inactivity.
How long things are kept
| The file you uploaded | 30 days, then deleted automatically. |
|---|---|
| Report, model, predictions | Kept. They are yours; no window runs against them. |
| De-identified data | 18 months or more, for model improvement, training and quality assurance. |
| Your email address | Until you delete your account. |
The window is a window on the upload. A result made from it outlives it by design: the report, the model and its evidence receipt are the thing you came for, and deleting them on a timer would delete your work rather than protect it. This page said the opposite until 31 August 2026 — that results were deleted with the data they came from — while the product had already stopped doing that, and a promise the product does not keep is worse than the one it does. If you would rather a result did not sit here, delete it: the control is on every session and it removes the bytes.
The original-upload window and result retention are enforced by the app, which serves
its retention policy at
https://cloud.veox.ai/api/retention — datasetDays
is the window above and artifactDays is null, which is what "kept"
means here. The de-identified-data policy is separate from that original-upload window.
De-identified data we keep
By uploading data through jAIn Free or Pro, including a customer-run workspace that sends data to our cloud, you agree that we may derive and retain de-identified data for model improvement, training and quality assurance. The default retention period is 18 months; de-identified data may be kept for longer. The original uploaded file is still deleted after 30 days.
Our initial process removes columns with recognized direct-identifier names and redacts cells matching selected patterns, such as email addresses, phone numbers, IPv4 addresses and card numbers. It does not find every identifier: identifying information can remain in column headings, free text, numeric values or combinations of otherwise ordinary fields. We may retain these copies before they have been checked against stronger validation criteria.
We handle these copies as sensitive data with restricted access; we do not sell or publish them. De-identification does not guarantee that a person cannot be identified. As we strengthen the methods and introduce stronger validation criteria, we will re-check earlier copies and delete those that fail the new criteria.
Deleting things
Every session has a delete control, and your account settings have a "delete my account and all data" control. These controls remove the stored account and session bytes, not just the listing. Account and session deletion is recorded so we can show it happened, but the record does not contain your data.
Deleting a table or your account does not remove de-identified copies we have already kept. To ask us to remove those, contact support@veox.ai. Models already trained with them are not retrained.
What we don't do
We do not sell your uploaded data or hand it to anyone else. We use it to produce your result and may use de-identified data for the purposes described above, including training shared models. Running the interface on your own machine does not mean that your data stays there.
Cookies
Every veox.ai site — these pages and the product at https://cloud.veox.ai — uses Google Analytics to count visits and see which parts get used. It is off until you say otherwise: the notice you were shown starts at no, so opting out is the state you are already in. Advertising features are switched off, so none of it builds an audience or follows you anywhere else.
There is a second one, and it is not a cookie. These pages are served by Cloudflare, and Cloudflare Web Analytics counts page views, referrers and how quickly pages render. It sets no cookie, stores nothing in your browser, and has no identifier for you — it cannot follow you between sites or between visits, which is why it is not behind the notice above and why there is nothing to opt out of. It runs on this site only; the product does not use it. We would rather name it than let you find it in a network log.
Your answer is one answer. It is kept in a cookie on veox.ai, so opting out here also opts you out inside the product and the other way round. It lives in your browser and nowhere else, which is why clearing your site data brings the notice back and why another device asks you again. Inside the product you can change it whenever you like, under Account → Cookies.
What the product reports is deliberately blunt: which screen was open, never which run. An
address like /runs/<your run id> is reported as /runs/:id, so
your run and dataset ids are not sent to Google at all.
Some cookies are not a choice, and we would rather say so than imply otherwise. Signing in sets a session cookie and a token that stops another site acting as you — that is what being signed in is, there is no version of the product without them, and they are never used to measure anything.
Some mail is the product working and arrives whatever you have said about marketing, because each of these is an answer to something you did:
- Sign-in links, which are how you get into your account
- The notice that one of your own runs has finished
- Anything we have to tell you about your account, your data or this service
Marketing email is anything we send you about Veox itself rather than about your own account: what we have built, what changed, and what we would like you to try. Nothing you do in the product turns it on — only the checkbox on your account page, or the one on our contact form — and turning it off never changes what the product does for you.
Ticking the box admits these three things and nothing else:
- What we have shipped, and what changed in the product
- Table shapes, models and features that are new enough to be worth a try
- The occasional question about what you need from Veox next
This is not our own description of the category: the app states it, unauthenticated, at https://cloud.veox.ai/api/marketing, and what you have just read is that statement rather than a paraphrase of it. It is the wording shown beside the checkbox itself, so the sentence you agreed to and the sentence quoted back at you are one sentence.
Saying yes, and taking it back
If you have an account, the checkbox is on your account page and works in both directions. If you do not, the contact form has its own, unticked: writing to us is not by itself consent to anything. Either way, every marketing message carries an unsubscribe link, pressing it is enough, and you do not need an account to use it.
Running the interface locally
The downloadable jAIn workspace runs its interface on your machine and sends your table to our cloud for computation. This policy applies to those uploads as well as uploads through the hosted app. A private deployment that keeps data entirely on your hardware is in design and is not shipped today.
Questions
support@veox.ai. A real person reads it.