Wants to look back at what data was sent to a model, and what came back.
Compliance · Prompt logging
Numaga records the full content of every AI interaction: the prompt, the context, the chosen model and the answer. Immutably kept, so you can see afterwards exactly what was asked and what came back.
Wants to look back at what data was sent to a model, and what came back.
Wants to reconstruct the exact interaction in the event of a complaint or incident, not just metadata.
Wants to investigate the full content on a suspicion of misuse.
The problem
Knowing that someone called a model is not enough. In the event of a data breach, a complaint or an inspection you want to know which data went in and what the model gave back, and none of that exists if all you keep is counts.
Numaga records the full content of every interaction: prompt, context, model and answer, immutably and with retention. What actually happened is retrievable, not reconstructed from fragments.
How it works
Record, retain, link, search.
For every interaction the prompt, context, model and answer are recorded: the complete exchange, not just a counter.
Logs are tamper-evident and kept to your retention term, up to seven years.
Every log is tied to the user, the guardrail hits and the risk class, so you see an interaction in its full context.
Search back by user, period or content, and export inspection-ready for an auditor or regulator.
In practice
Numaga is running a pilot with Ontdek Zorg, a home-care organisation in Gelderland. Healthcare, with special-category personal data under the GDPR, is where responsible AI is hardest. That is exactly where we put the control plane to the test.
We show a full interaction log and an inspection-ready export in a demo environment.
Request demo access