- Documentor
- Data security
Documents under lock and key. A safe second copy. Data where you want it.
A documentation archive is often the most valuable thing an organisation owns – contracts, drawings, inspection reports, accounts. That is why Documentor encrypts an archive’s files with a key that belongs to that archive alone, backs them up, encrypted, to separate storage and lets you choose where they are kept. Everything is switched on in the archive settings, and the archive can be read and searched while it is being switched on.
What protects your documents.
Each layer covers a different risk: encryption covers files leaking from storage, backups cover losing them, and your own storage settles who physically holds the data. They are switched on per archive, so sensitive archives can be protected more than the rest.
Unreadable without the key
Documentor encrypts every file before it stores it. The storage only ever receives ciphertext – whether it is your server, a cloud or our storage.
- AES-256-GCM, a separate key for every archive
- File name and sender are encrypted too
- Switched on in the background, no downtime
A copy elsewhere, encrypted
The archive is backed up regularly to separate storage, for example Backblaze B2. The backup provider cannot read anything in it.
- Originals, document data and AI results
- Only changes are uploaded
- An email when the backup keeps failing
Files where you want them
Every group can have its own S3-compatible storage – on your premises, in a cloud or with us. An archive can be moved to other storage while in use.
- Access keys stored encrypted only
- Changes approved by a supervisor
- Never silently somewhere else
Not even the storage operator can read the content.
Documentor encrypts a file itself before it leaves for storage, and decrypts it only when a user with access to the archive, or processing, needs it. Search, page previews and downloads work exactly as they do without encryption.
- Switching ona supervisor or group administrator in the archive settings, Storage panel
- While in useexisting files are encrypted in the background; the archive can be read and searched meanwhile, new documents wait and are processed right after
- What is encryptedoriginals, page images and previews, searchable PDFs, and the file name and sender stored with the file
- No exceptionsnothing is ever written to an encrypted archive unencrypted – if the key is missing, the application reports an error
- Tamper-proofa truncated, reordered, substituted or renamed file is rejected
- Switching offencryption can be switched off again; the files are decrypted in the background the same way
- Cipher
- AES-256-GCM in 64 KiB blocks, each block authenticated
- Archive key
- random, a separate one for every archive
- File key
- derived for every file separately (HKDF-SHA256)
- Master key
- protects the archive keys; kept neither in the database nor in storage
- Key rotation
- the master key can be replaced without re-encrypting files
- Overhead
- a file grows by roughly 0.02 %
- Object names
- a content hash or document number, never its title
Passwords and keys encrypted only
Documentor stores storage access keys and AI model API keys encrypted with a key that is not in the database. Not even a copy of the database gives them away, and once saved the application never shows them to anyone again.
What file encryption does not cover
The recognised text of documents and the search index live in the database and the index, because search and chat work from them. They are protected by the application’s access rights; if you want them on your premises too, choose the deployment entirely on your premises.
A second copy elsewhere, encrypted and monitored.
An archive’s backup goes to different storage from the archive itself. It is switched on by a supervisor or group administrator in the archive settings (Backup panel), because the group pays for it.
One failure cannot take both copies
The target is S3-compatible storage separate from the archive’s storage, for example Backblaze B2. It can be our backup storage or your own – it is set in the group configuration.
Encrypted before it leaves
Documentor encrypts every backup file (AES-256-GCM) while it is still on the server. The provider sees only sizes and technical file names; the backup key is not in the database.
You know it is running
The Backup panel shows the last backup, its size and its price. If the backup keeps failing, the group administrator gets an email; when we host Documentor, our monitoring watches it too.
- What the backup contains
- Originalsevery document exactly as it arrived
- Converted PDFsfrom Word, images, Markdown and other formats
- Document dataannotations, labels, codes and AI descriptions of images and charts – no need to compute them again
- Archive settingsinstructions for the AI, labels, code rules and models
- Nothing redundantpage images, OCR and the index are rebuilt from the originals, so they are not backed up
- How it runs
- Incrementalonly new and changed files are uploaded, nothing twice
- Deleted versionsa deleted or overwritten file is kept by the backup storage for the retention period set on the target
- Encrypted archivethe backup has its own key, independent of the archive’s keys
- Pricebackups to our storage are charged by volume, backups to your own are not; the volume is shown in usage and in the plan consumption
And operational backups underneath
Archive backups are a service for you. Independently of them, when Documentor runs in our cluster we back up the database continuously and every day copy the storage and the database backups, encrypted, to a data centre in the EU outside our infrastructure, with versions protected against deletion.
You decide who holds your data.
An archive’s files – originals, page images and searchable PDFs – are kept in S3-compatible storage. Every group can have its own: a server on your network, a cloud of your choice or our storage. Where the rest lives depends on the deployment option.
| What | Entirely on your premises | Your data, our operations | Entirely with us |
|---|---|---|---|
| originals, page images, searchable PDFs | your storage | your storage | our storage, or yours for selected groups |
| database with the recognised text | with you | with us | with us |
| search index | with you | with us | with us |
| AI models | on your server | on our server, or yours | on our server, or yours |
| external third-party AI models | only with the archive’s consent | only with the archive’s consent | only with the archive’s consent |
| archive backups | wherever you choose | wherever you choose | our backup storage, or yours |
Nobody sees the keys
Storage connections belong to the group; an archive only picks from what is on offer. The access key is stored encrypted and the application never shows it to anyone again once saved.
New storage only once approved
A group administrator proposes a connection; a supervisor tests it and approves it or rejects it with a reason. Until then the existing settings apply.
To other storage while in use
An archive can be moved elsewhere; the application checks that every file is on the target before it switches over. The original data stays until an administrator deletes it.
An outage is an error, not a fallback
If your storage is unavailable, the application says so. It never quietly saves files to any server other than the archive’s own.
Every document only for those with access to it.
Encryption protects files in storage; inside the application, roles protect them. Who can do what is described on the Groups and access page.
No passwords of its own
You sign in with a Microsoft or Google account; Documentor keeps no passwords. A new user sees nothing until a supervisor approves them or an Archive administrator gives them a role.
Every archive on its own
An archive has its own place in storage and its own index. Search, chat and research only work with the archives the user has access to – even when searching several at once.
Reader, Processor, Administrator
Roles are assigned per archive, to people and to groups. A blocked user has no access even through a group, and a supervisor cannot remove their own role or block themselves.
Only with the archive’s consent
A third-party model receives document content only from an archive that has explicitly allowed it. All other archives use only models that do not run at a third party.
Text recognition stays at home
Text recognition, page layout and reading QR codes and barcodes run inside Documentor. No external service needs to see a single page for that.
You know what is going on
Usage shows who searched and asked the AI and how many tokens went to external models, separately from internal ones. Processing and backup errors are reported by email notifications.
We will show you how it works on your own documents.
Send us a few PDFs from your archive – old scans are fine. In 30 minutes you will see how searching them works and how the chat answers.
Already a customer? Sign in