Protect
what you
owe them.

Write your question with real client names, case numbers, and all the context the AI needs. Or upload a PDF, Word document, or text file — Naosuu extracts the text automatically and adds it to your prompt, ready for redaction. Note: document upload works with text-based files. Naosuu extracts the text layer — it does not process scanned images as images.

Before your prompt goes anywhere, Naosuu opens a preflight panel showing you exactly what it detected. Every identified element is listed — you can check or uncheck each item individually. If Naosuu missed something, you can add words or phrases manually at the bottom of the panel. Nothing is sent until you approve it. You are always the last line of defence.

Once you confirm, Naosuu replaces every approved item with an anonymous placeholder and sends the clean prompt to the AI. No names. No case numbers. No identifying information of any kind — just context. And critically: no AI is involved in this process. Naosuu uses rule-based pattern matching and named entity recognition running entirely on your machine. Your data is never processed by a model, never sent to Naosuu, never seen by anyone but you.

When the AI responds, Naosuu swaps every placeholder back to the real name or value — locally, in your browser, before the text appears on screen. You get a response that reads as if the AI knew everything. It never did.