Use the best AI models while client details stay in your office.
The Privacy Gateway sits on your own server, between your team and any cloud AI model. It removes clients’ personal information before a request leaves the office, and restores it when the answer comes back.
Draft a reply to Ana Ferreira confirming we have her passport PA4417302 for matter M-20391.
Draft a reply to [PERSON_K7Q2] confirming we have her passport [PASSPORT_9D1M] for matter [MATTER_4HX8].
Dear Ana Ferreira, we have received your passport and added it to your file for matter M-20391. We’ll be in touch once your nomination is lodged.
Four steps on every request.
Detect
A model running on your own server finds the personal information in the request: names, passport numbers, email addresses and matter codes.
Replace
Each detail is swapped for a placeholder such as [PERSON_K7Q2]. The meaning of the request stays intact, so the AI model can still do the work.
Send
Only the text with placeholders leaves your office for the cloud model.
Restore
The answer comes back with placeholders, and the gateway puts the real details back before your staff read it.
Your staff can see exactly what was sent.
For any request, staff can compare what was typed in the office with what the AI model received. When a client asks how their information is handled, you can show them.
Private work stays local. Hard questions get the best model.
Local models on your own server answer everyday questions from your files and pick out personal information, so that work never leaves the building.
For research and harder questions, such as drafting a case story or summarising a long file, the gateway lets your team use frontier reasoning models, with personal information removed first.