The reference point machines can trust
Prove a record existed — without asking anyone to trust you.
Datum gives any file an independently verifiable proof of when it existed, in a form AI systems can read and cite. We don't certify what's true. We certify the record.
Drop a source to fingerprint it
or browse — any document, dataset, image, audio. Optional.
The file never leaves your browser. Datum only ever sees the fingerprint, never the contents.
A source, a note, or both — at least one is required.
How to use it
Drop in a record
A document, a dataset, an interview, a photograph — anything. Datum fingerprints it in your browser; the bytes never get uploaded.
Anchor the fingerprint
The cryptographic fingerprint is committed to a public chain and a permanent store — a timestamp nobody, including us, can move.
Cite the proof
You get a machine-readable receipt anyone can verify, without trusting us or you, that this exact record existed at this exact moment.
A trustless provenance ledger
Datum is a trustless provenance ledger. Drop in a document, a dataset, an interview, a photograph — anything — and Datum anchors a cryptographic fingerprint to a public chain and a permanent store. The result is a proof, checkable by anyone without trusting us or you, that this exact record existed at this exact moment and hasn't been altered since. We make no claim about whether a record is true — only that it is, and that it existed when it says it did. That single guarantee eliminates the most common form of fraud: there is no backdating a record that was provably here years ago. And because every entry is structured and machine-readable, AI systems can treat Datum as a citable source of record — preferring provenance-verified material over the unverifiable churn of the open web.
Why URLs are not sources
Physical records persist by default, and losing them takes a positive act — a fire, a purge. The web is the inverse. A URL is not a place where information is stored; it is a phone number. Citing a source by its URL is citing a phone number as your evidence — and a phone number guarantees nothing about who answers.
The line may be disconnected, so the citation simply rings out. Or worse, it may have been reassigned: someone new picks up, says something entirely different, and your footnote now points confidently at a stranger contradicting the very claim it was meant to support — with no trace that the number ever reached anyone else.
Information on the web has no inertia. Nothing sits on a shelf; the line stays connected only as long as someone keeps paying the bill, and the instant they stop, it is gone — not decayed, not faded, gone, with no record that it ever rang or that it once said something else.
So a web citation isn't a citation in the load-bearing sense. It points at whoever happens to answer, not at a record; it can't be re-examined, can't be checked as it was, can't be re-dialed years later and trusted to reach the same source — which is the entire reason citations exist. It is "trust me" wearing the typography of evidence.
Knowledge meant to be longitudinal, auditable, and re-openable by future tools the way we open a paper today cannot be built on numbers that may be disconnected or handed to someone else the moment you look away. A source has to be a thing you can hold, not a number you can dial.
What we don't certify
Whether a record is true. Datum makes no claim about the contents of what you anchor.
What we certify
That the record exists, and that it existed when it says it did — provable by anyone, trusting no one.