Share documents and use AI, without the sensitive details
How El Niño built ShareSafe: a product that replaces names, amounts and other sensitive details in documents with neutral placeholders, so teams can share them and use them with AI.
- languages
- 5
- certified
- ISO 27001

Confidential documents and AI don't mix easily
Sensitive details everywhere
Solicitors, accountants, schools and healthcare teams work with confidential documents every day. They want AI to summarise a contract, or an adviser to look at a case file. But the names, addresses, rates and account numbers in those documents don't need to go along.
Removing them by hand doesn't scale
Blacking out details by hand takes time and is easy to get wrong. And what has to stay hidden differs per task: for a financial analysis the figures matter, for a review of notice terms they don't.
So we built ShareSafe
ShareSafe is El Niño's own product. It finds sensitive details in a document and replaces them with neutral placeholders, such as [NAME_01] for a person or [ORG_01] for an organisation. You check the protected version, adjust it where needed, and then share it or use it with AI. Placeholders can be reversible, so the real details can be put back later, or permanent.
![A draft claim form in ShareSafe: the original with names, addresses and amounts, and the protected version with placeholders such as [NAME_01] (fictional data)](/_next/image?url=https%3A%2F%2Fcms.elnino.tech%2Fassets%2Fcases%2Fcase-sharesafe-share-documents-without-the-sensitive-details%2Fsharesafe-original-and-protected.png&w=3840&q=75)
One product, four ways to use it
ShareSafe Free
Protect one file without an account. ShareSafe doesn't keep the file or the recovery key after processing.
ShareSafe Workspace
Originals and protected versions together, stored in the EU. Teams decide who may see the original details, share files through links that expire, and can withdraw access again. An activity log shows what happened to every file.
ShareSafe in Claude and ChatGPT
Connect the Workspace to Claude or ChatGPT, and the AI searches, compares and summarises the protected documents. Want to know who is behind [NAME_01]? You look it up in a private ShareSafe panel, which is designed to keep that name out of the chat.
AI Safety Assistant
A free Chrome extension that works like a spellchecker for confidential information. While you write a prompt in ChatGPT, Claude, Gemini or Perplexity, it points out sensitive details and replaces them in one click. Passwords, API keys and tokens are recognised on your own device.

How we built it
Protection first
Every document goes through the same steps: a virus scan, text extraction, detection of sensitive details and replacement with placeholders. If the required protection can't be applied, the file isn't passed on. The link between placeholders and original values is stored encrypted.
Built for teams
Each organisation's vaults are strictly separated, with access rights per vault and a complete audit trail. The backend is built in Python with FastAPI, with Celery workers for the heavy processing. The web app is built with Next.js.
The AI models
Sensitive details are found by GLiNER, a multilingual model that recognises names, organisations, addresses and other entities, together with rules for fixed patterns such as IBANs and email addresses. To search documents, ShareSafe uses Qwen3 Embedding, and smaller tasks like summaries and titles are done by Qwen3 4B Instruct. These models run on El Niño's own infrastructure, so sensitive details are found and replaced before anything reaches an outside AI provider. For answering questions, an organisation connects the AI provider of its choice with its own key, or works from Claude or ChatGPT. Whichever model answers only ever sees the protected version.
Connected to the AI tools people already use
The Claude and ChatGPT connections are built on the Model Context Protocol (MCP), with ShareSafe handling the sign-in. Vaults reach the AI app only under short neutral names, and only for the people who have access.
Built and operated from the EU
ShareSafe runs on our own Kubernetes infrastructure in the EU and is ISO 27001 certified.
The result
ShareSafe is live on sharesafe.ai, in English, Dutch, German, French and Spanish, with a free 30-day trial. Teams can now use AI on contracts, case files and pupil or patient records, while the sensitive details stay with them. And what we learn building ShareSafe goes straight into the AI solutions we build for clients.


