What was used
Trace records sources, document versions, the model and technical parameters such as query time and cost.
Auditable AI
Every AI answer has a source, a trace and a control result before delivery.
QTJ records the answer context and checks its quality before the user sees it. Trace and Judge create evidence instead of relying on a claim that the model should work correctly.
A trace for every answerJudge before deliveryEvidence Pack for audit
QTJ · Query · Trace · Judge
Auditable AI
Auditable AI answers with answer ID, sources, document versions, model, cost, time and quality result in an Evidence Pack.
QTJ.ai · BREPO Sp. z o.o.Trace: from an answer to its technical basis.
Every result receives an identifier and a dataset that allows its basis to be reconstructed.
What the user received
The answer_id connects the delivered answer with its context, validation and source materials.
Judge: AI answer quality control.
Validators run before delivery and can return PASS, WARNING, REFUSAL or ESCALATION.
Answer risks
Control can cover source grounding, hallucinations, sensitive data, company policies and procedural execution risk.
Evidence Pack
Trace and Judge results feed a report that organises material for internal review or audit.
Common questions
Does auditability mean automatic legal compliance?
No. QTJ provides technical mechanisms and evidence; legal obligations depend on the specific system, role and use case.
Does control run after generation?
Judge checks the model output before it is shown to the user and can warn, refuse or escalate it.
Does the trace remain after changing the model?
Yes. The model is replaceable while logging and the audit layer remain part of QTJ.
QTJ.ai
Book a pilot
QTJ records the answer context and checks its quality before the user sees it. Trace and Judge create evidence instead of relying on a claim that the model should work correctly.
Book a pilot- E-mailqtj@qtj.ai
- BREPOMikołów · Poland