The working student
What does he find?
What his group released. Nothing else.
Responsible AI
An answer is only worth as much as its origin. So we build systems that back up their answers. And respect the boundaries you set.
Why it matters
If teams are to trust an AI, they must be able to check its answers. One wrong figure in a quote or one invented reference in a report costs more than the AI ever saves. Verifiability is the condition under which AI belongs in a company.
How long is filter F-203 approved?
The approval for filter F-203 runs until 30 June 2027. It is checked every six months according to maintenance plan W-12; maintenance is in charge.
The check lives in the same database query that selects the results. A passage without clearance is missing from the result itself. The model never gets to read it. And what was never read can never surface in a summary.
Language models run locally on your hardware, with open weights. Your questions never leave the company to get answered. And your content trains no external model.
What the agent did remains as a logged run: which sources it pulled, which tools it used. You can trace it step by step.
Everything essential on the GDPR, the EU AI Act, and storage and logging questions is on our compliance page. Concrete enough to hold against your own question catalogue.
What a search finds depends on who asks. That is enforced in the query itself.
What does he find?
What his group released. Nothing else.
And if he asks about a colleague’s salary?
He gets no answer, because there is no hit.
How does the system know what she may see?
From your existing groups.
Can she verify this?
Yes, without asking us.
Every answer lays its sources open.
We go through them concretely, line by line. With your IT leadership, your data protection officer or your works council.