The AI you already reach for — now on your own terms.
Most people already use a chatbot every day — the real question is which one, and what happens to your data. There are four options, and three of them come with a catch.
01
Don't use AI
A real choice — though most teams use one anyway, from a personal account the company can't see.
02
An American assistant
Capable and familiar — though the Cloud Act lets US authorities request what's stored with a US-controlled provider, wherever it sits.
03
Build your own
Starting is straightforward. Picking the right model every week and keeping it running is the ongoing part.
04
A European assistant
GDPRchat is EU-owned and EU-operated — FRITS AI keeps it current, and the transparency page shows exactly what runs.
Who's it for?
Swap today — no strings attached.
The full assistant in every official EU language, with the whole shelf of tools. Free to start, no card, and your conversations are processed and stored only in Europe.
One plan covers everyone and one bill covers the lot. Your email domain is what lets colleagues in, so nobody creates accounts for anybody — no IT project, no rollout quarter. It is running today.
A pre-signed data processing agreement, a sub-processor list you can hand to your DPO, and invoicing by EAN — because a public body cannot pay by card. We set that up with you directly.
What comes next Tools for the team is where four hundred contracts get asked one question at once, together.
Tools for the team
This is where it starts working with your own material — your documents, your rules, your team.
A chatbot answers general questions well. These tools go further — the answer comes out of your own material, and your DPO can see exactly where it went.
Ask PDFs
Drop in a 200-page tender and ask what the penalty clause says.
Document AI
Run the same question across four hundred contracts at once.
Bots
Answer the questions your own documents already answer — without a person in the loop.
Workplace
Add the whole team, and decide who can see what.
A pre-signed DPA
Ready before your DPO asks for it, with the sub-processor list beside it.
One shared pool
Every tool included for everybody. Nothing held back for a bigger plan.
Where should we start?
Start with the pack, not the pitch.
The data processing agreement is pre-signed and public. So is the list of every sub-processor, and the page that says which model handles which job and where it physically runs. Your DPO can read all of it before you ever speak to us.
Operations, security, updates — and somebody deciding every week which model is the right one for each job. That last part is what a self-built setup usually needs the most ongoing help with, and it is the thing we keep doing. You can check the current answer at any time.
One plan, one bill, and one pool of credits shared by the whole company — so the colleague who barely uses it never strands an allowance, and somebody's heavy week is absorbed by the team. Volume pricing, annual invoicing and EAN exist as a door, not as a toll gate.
What comes next The API is where it stops needing a person to ask — running inside your own systems, day and night.
The API
This is where it becomes part of how your own systems already run.
Our endpoint is OpenAI-compatible. Swap the base URL and the key, change nothing else, and the same code runs on European models, in Europe — whenever you're ready.
base_url = "https://gdprchat.eu/api/v1" # the only line that changes
api_key = "sk-..." # made in your own account, now
This is where the gain stops being a percentage: work that needed a person now happens on its own.
Where should it run?
Then go and make one.
Self-serve. Create a key in your account and start — no sales call, no minimum, and the same one bill as the rest of the company's usage.
We install and run the whole thing on your own hardware, on your own network. Same product, same tools, inside your building — for the organisations where "in Europe" is not close enough.
Automation and agentic-AI workshops for the people who decide, drawing on what we learn running European AI in production — including the parts that did not work.
What we owe you here From here you're the one building. What we owe you is honest measurements, published — including the ones that didn't go our way. The technical reports →
Under the US Cloud Act, American authorities can compel a US company to hand over data it holds — wherever in the world it sits. European hosting does not change it. A data processing agreement does not change it. It follows the company, not the server. FRITS AI ApS is Danish, the models are European, the hosting is European — and we publish which model does which job, so you can check it rather than believe it.
Running European AI in production keeps surfacing problems nobody has written down — including what right-sizing a model actually saves in energy. We publish them as numbered, citable technical reports, measurements and failures included. Self-published, not peer-reviewed, and we say so.
