Chat withChatGPT,
now through Minac.
Cloud answers. Private data.
A private AI gateway that strips personal data out of every cloud-bound prompt and restores it in the response — so your team gets the best cloud models while sensitive data stays inside your walls.
Curious right now? See the demo playground
the basics
What is PII — and how Minac handles it.
PII (personally identifiable information) is any detail that can identify a real person — on its own or combined with other details. Minac detects it on your own hardware, replaces each value with a typed placeholder before the prompt leaves your network, and restores the originals in the reply. The cloud model only ever sees the placeholder.
detected locally · replaced before egress · restored on return
what minac does
Frontier models, on your terms.
Minac sits between your people and the cloud LLM. The model sees your full prompt with every sensitive value swapped for a placeholder; your data never leaves the building intact.
A private AI gateway is a layer between your team and a cloud language model: it removes sensitive data from a prompt before the prompt is sent, and restores that data in the answer. It lets an organization use frontier cloud models without exposing personal or confidential information to the provider.
Stripped on egress
Every cloud-bound prompt is scanned and personal values are swapped for placeholders before anything leaves your network.
Returned intact
When the answer comes back, Minac restores your original values — and blocks any new PII the model tries to invent.
An audit trail
Compliance and security teams get a quiet, factual record of what was protected, without the security theatre.
Minac enforces all three steps automatically. On every outbound prompt it detects sensitive values and substitutes typed placeholders, then sends the sanitized prompt to the model. When the answer returns, it restores the original values before your team reads the reply, and holds back any new personal data the model tries to invent.
security model
Your data stays in your environment.
Minac runs on infrastructure you already own. Sensitive values are detected there and replaced before any prompt reaches the cloud model; the originals are restored locally in the reply. The provider works on the sanitized prompt and never receives the sensitive data.
- Detected on your infrastructureMulti-layer detection runs locally — names, emails, IDs, account details, secrets, and your own custom terms are found before anything is sent.
- Replaced with placeholdersThe cloud model receives the prompt with each sensitive value swapped for a typed placeholder, so it keeps enough context to answer without seeing the original.
- Restored on returnOriginal values are restored locally, before your team reads the reply.
- Invented data blockedThe model's response is re-scanned, and any new personal data it invents is kept redacted.
- Review when it mattersIn Review mode, users can approve detected values or tag extra sensitive phrases by hand before sending.
- An audit trail you controlA factual record of what was protected stays inside your environment — 30 days by default, adjustable per customer. Raw sensitive values are never stored in external systems.
security questions or disclosures · [email protected]
the signature
You see your data. The cloud doesn't.
The privacy mechanism is made visible — detected values are underlined, named, and replaced. Nothing dramatic; just honest.
Dana Whitfield → [NAME] · [email protected] → [EMAIL]
Only the placeholder text reaches the cloud model.
Your values return in the reply; you never lost context.
In this example, the prompt is sanitized before it leaves: the name becomes [NAME] and the address becomes [EMAIL]. The cloud model receives the full instruction with only those two values replaced, so it can still write the reply. When the response returns, Minac restores Dana Whitfield and [email protected] in place of the placeholders.
voice in, safely
Dictate and attach audio — still private.
Voice notes and recordings are transcribed on your hardware, then the transcript runs through the same sanitizer before any of it reaches the cloud.
Call Dana Whitfield back about account GB29 NWBK 6016 before Friday.
documents & images
PII redacted in the pixels, not just the text.
Attach a scan, a photo, or a PDF. Minac strips embedded metadata, reads the page with on-device OCR, and black-boxes any detected personal data before the file is forwarded.
From:
Sent:
To:
Subject: Prompt test data for HR assistant pilot
Hi ,
I wanted to share a quick update on the HR assistant pilot and the prompt test data we plan to use next week.
The pilot will start with onboarding, payroll, and internal support scenarios. We need to confirm the assistant can identify and handle personal information correctly, especially when the data includes Finnish characters such as ä and ö.
Prompt 1:
"Create an onboarding checklist for employee employee ID . Her private email is"
Prompt 2:
Prompt 3:
"Draft a support reply to about her locked HR account. Include her last login IP , recovery phone , and office location"
Best regards,
HR Systems Manager, ACME HR Solutions
Mobile:
redacted copy saved · original never leaves your network
three modes
As much control as the moment needs.
Pick how visible the sanitizer is for each message.
Detect and replace PII silently on every message.
Review what will be sanitized before each send.
Sanitization disabled. Messages are sent as written — for non-sensitive work.
review mode
Tag anything as sensitive — not just the obvious.
Auto-detection catches names, emails, and IDs. In Review mode you can highlight anything else — a project name, a client, a phrase — and tag it so the cloud sees a placeholder in its place.
questions
Common questions.
What is a private AI gateway?
A private AI gateway is a layer between your team and cloud LLMs that keeps sensitive data out of the provider's hands. Minac detects personal and confidential values locally, replaces each one with a typed placeholder before the prompt leaves your environment, sends the sanitized prompt to the model, and restores the originals in the reply — so you get frontier-model answers without handing over the underlying data.
Does Minac work with ChatGPT, Claude, and Gemini?
Yes. Minac connects to leading cloud models such as ChatGPT, Claude, and Gemini, so your team keeps using the model it already prefers. Whichever you choose, the prompt is sanitized the same way before it leaves: sensitive values are replaced with placeholders, and your sensitive data never reaches the provider in its original form.
How does Minac handle non-English prompts?
Detection strength is even across major European and Nordic languages — it isn't tuned to English. German, Spanish, French, Italian, Dutch, and the Nordic languages all hold up in testing. Finnish has been tested most extensively, including informal spoken and genitive name forms, and detection tolerates typos and obfuscated spellings such as name(at)company(dot)fi, not just clean text.
What does the cloud provider see, and where is my data stored?
The provider receives your full prompt, but every sensitive value — dictionary terms, values Minac detects, and anything you tag in review — is swapped for a typed placeholder first, so it never sees the originals. Minac runs on infrastructure you already own, for example a dedicated virtual machine in your own Azure tenant, and your conversations stay inside that environment — retained for purposes such as audit logging and GDPR accountability, not sent anywhere else.
Who is Minac for?
Minac is built for the whole organization, not a single team. Anyone can use it like the everyday cloud model they already reach for, across every department rather than a walled-off pilot. It's especially useful for decision-makers who want a private sounding board on sensitive or early-stage work. The ideal deployment is company-wide, so protection is the default for everyone.
Can Minac protect our company's own terms, not just personal data?
Yes. Alongside standard personal data, Minac keeps a customer dictionary of terms specific to you — project names, client names, even your own company name — and redacts them before prompts reach the cloud. Beyond protecting each value, this makes it harder for a provider to link prompts together and infer who you are or what you're working on.
stay in the loop
Register your interest.
we'll only use this to tell you about minac.
noted — we'll be in touch.