Kahubi for journal editors

An AI first read of every submission. The manuscript is never kept.

Editors are trusted with ideas that are not published yet. Sealed review lets you use AI on a submission without that trust leaking: the manuscript is read in memory and never saved, only AI hosted in the EU reads it, and the review is encrypted and destroyed when you are done. Every deletion leaves a receipt you can file or show the author.

Free plan included. No credit card required.

Where Kahubi fits your work

Sealed review

Drop the PDF, get a desk assessment

Drop the submission as PDF or Word. In about a minute you get a suggested decision (send out, return to the authors, desk reject) with its reasons, the fit with your journal, and the issues the authors need to address.

Screening checklist

The checks nobody enjoys, done first

Ethics approval and consent, data availability, funding and conflicts of interest, anonymisation for blind review, length and structure: each marked OK, concern or missing with a one-line note.

Reviewers

Know which reviewers to look for

The assessment names the expertise the peer reviewers need, specific enough to search your reviewer pool. It never suggests named people.

Referee report

Or a referee report for the authors

Switch to a referee report: summary, recommendation, major and minor comments. Copy it straight into your journal system, with a separate copy that leaves out your confidential notes, or download it as Word.

Destroyed on schedule

Encrypted, then crypto-shredded

The review is stored encrypted with a key of its own, kept outside the database and all backups. Pick when it is destroyed: right after you download it, after 24 hours, 7 days or 30 days. When the key is gone, no copy anywhere can be read again, not even by us.

Editorial boards

One switch for the whole board

The board owner turns on "Require sealed reviews" in Team. From then on every member’s peer reviews and similarity checks run sealed, and nobody can turn it off by mistake.

AI chat

Chat with your library, not a generic chatbot

Kahubi answers from the papers, drafts, datasets, and transcripts you actually have. Type @ to pull any source into the conversation; the agent reads it, searches your library, and cites what it used.

  • @-mention papers, drafts, datasets, interviews, and files
  • Select text in any PDF to explain, summarize, or translate it
  • Agent searches OpenAlex and the web, and imports papers for you
Try it free
Kahubi's AI chat answering from the user's library, with a tool chip and cited references
Writing

Write accurate manuscripts, grounded in your sources

Draft in a clean editor and let the AI help — it writes in your voice, learned from your own publications, and every paragraph stays anchored to the papers, data, and notes you actually have, never invented. Read it as a formatted PDF, edit it like a Word document, or drop into LaTeX when you want fine control.

  • Writes in your voice: a style profile learned from your published work, not a generic chatbot register
  • Grounded in your library: the AI writes from your real papers and data, and cites what it used
  • One click to a compiled PDF, plus Word (.docx) export and full version history
Start a draft
A manuscript compiled to a typeset PDF in Kahubi

What changes

  • The manuscript is never stored: not in a library, a project, the database or a backup
  • EU only: Kahubi runs in Sweden, the AI on EU providers (Ireland, Sweden, the Netherlands); if none is available the review fails instead of leaving the EU
  • Every destroyed review leaves a PDF receipt with the manuscript’s fingerprint, never its content
  • A sealed similarity check on the same manuscript in one click, also EU only
  • Content is not used to train AI models, and a DPA is available for journals and publishers

Common questions

Where exactly does the manuscript go?
Your browser sends the file to Kahubi’s servers in Sweden, which read the text in memory and send it back to your browser tab. When you start the review, the text goes once more through our servers to an EU-hosted AI model (TensorX in Ireland, with Inceptron and Berget AI in Sweden or GreenPT in the Netherlands as backups). Nothing is written to disk or the database on the way. A sealed review refuses any AI provider outside the EU, even if your account is set up with a different one.
What is actually stored, and for how long?
Only the AI review, encrypted with a key that is kept outside the database and outside all backups, until the time you picked: right after you download it, 24 hours, 7 days or 30 days. Then the key is overwritten and deleted. Encrypted copies that remain in database backups can no longer be read by anyone. What we keep is the receipt: when, how many words, the manuscript’s SHA-256 fingerprint and where it was processed.
What can you not promise?
While the AI works, usually under a minute, the text is in the memory of our server and of the EU model provider, which processes it under a data processing agreement and may not use it for training. Files you download and text you copy are in your hands. And the AI can be wrong: it is a first read, and the decision stays with you.
Can our whole editorial board use it the same way?
Yes. Create a team, invite the editors, and the owner turns on "Require sealed reviews" and picks the default deletion time. Every member’s reviews and similarity checks are then sealed, enforced on our servers, not just in the interface.
Do we need a contract?
Individual editors can start on the free plan. For a journal or publisher we sign a data processing agreement, and invoices work for institutional customers.

Read the next submission with a sealed review

Free to start. The manuscript is never kept, and you get a receipt when the review is destroyed.

Kahubi for other roles