Privacy notice

Where your data goes

The short answer is nowhere. There is no account to create, no email to give us, no form to fill in, no document to upload and no scan to run — this site has no way for you to send us anything about yourself. What follows says what that means in practice, what Datum does hold, and how to check both without taking our word for it.

How the site is built

Datum is a read-only reference for AI news, models and measurements. Every page is generated from a database of public material — publishers' feeds, evaluators' published scores and sellers' own price pages — and rendered on the server. Nothing on it accepts input.

• ·No account, no sign-up, no email. There is no login and no mailing list, and nothing on the site asks who you are.
• ·No forms at all. No search box that posts answers, no newsletter field, no upload — and no API route that accepts anything from a browser.
• ·No cookies, and nothing in your browser's storage. The site sets no cookie and writes nothing to local or session storage, so there is nothing here to clear and nothing to consent to.
• ·No advertising and no third-party trackers. No ad code, no social pixel, no cross-site identifier.

What the server unavoidably sees

Every website is served by a machine, and the machine sees a request. Our hosting and database providers record ordinary technical logs — IP address, time, requested path, user agent — to keep the service secure and running, as every host does. We do not use them to build a profile, we do not join them to anything, and we do not sell or share them for marketing. They sit under the host's retention and are not part of Datum's own database.

Full disclosure

Datum loads no analytics script today. The codebase carries a placeholder for aggregate, cookieless page analytics; it is unset, and nothing is collected. If that changes, this section names it before it is switched on: it would record a page, a country and a referrer in aggregate, with no cookie, no identifier, and nothing that follows you to another site.

What Datum does hold

The database holds public material about AI. None of it is about you, and the retention rules below are enforced by the pipeline rather than promised — the status board shows when each stage, including the retention job, last ran.

• ·Stories and sources— our own headlines and summaries, plus a link to each publisher's article. Publisher body text is stored briefly so it can be classified and summarised, and is deleted seven days after fetch. Our own article and story rows are permanent, so every story URL keeps working.
• ·Embeddings — the vectors used to group stories are retired after 30 days.
• ·Rejected press articles — deleted after 90 days. Nothing else is ever deleted.
• ·Measurements— benchmark scores with their evaluation date and scaffold, from Epoch AI's CC-BY-4.0 index and the evaluators' own pages.
• ·Prices— the sellers' own published rates, and the CC-BY-4.0 AI Price Index for sellers we may not read directly.

The crawler

DatumBot polls public feeds and a small number of published pages to build the news wire. What it fetches, how fast, and how to stop it is on the crawler policy page. It honours robots.txt, prefers the feed to the page, takes the longer of our own rate and any Crawl-delay, never uses an archive mirror to get around a paywall, and does not crawl for training data. If you want your source removed, ask and we remove it — no justification required.

How to check all of this without asking us

Any claim of this kind is worth what it can be tested against. Four checks, and the first needs nothing from us.

• ·Open the network tab. Your browser's developer tools show every request this site makes. Load a page and follow it: no request carries anything you entered, because there is nowhere to enter it.
• ·Look for cookies. The storage panel is empty for this site.
• ·Read the crawler policy. It is the page the URL in our User-Agent points at, and it describes the fetches we do make as well as the ones we do not.
• ·Watch the status board. It reports, stage by stage, when the pipeline last ran — including the retention job that deletes publisher text.

Your rights

Under UK GDPR you can ask to access, correct or delete personal data, to restrict or object to certain processing, and to receive a copy of data you gave us. You can also complain to the Information Commissioner's Office at ico.org.uk. Because the site holds nothing about its readers, there is usually nothing to access or delete — but if you have written to us and want that correspondence dealt with, ask and we will respond within one month.

Contact

For any privacy question, email privacy@thedatum.uk. For anything else, io@thedatum.uk.

We may update this notice; the date at the top shows when it last changed.