Audit & Explainability
Every important output records its source, model/version, timestamp, user, correction and approval outcome.
Synthetic demonstration
Audit log (2)
| Time | User / role | Action | Target | Source | Model | Outcome |
|---|---|---|---|---|---|---|
| 10/28/2026, 9:12:00 AM | S. Prakash (Investigator) investigator | Extraction corrected | C-2025-0042 / accused name | D3 p1¶2 | extract-v1.4 | pending review |
| 10/27/2026, 3:40:00 PM | M. Reddy (Supervisor) supervisor | Legal suggestion accepted | C-2024-0117 / Sec. 34(b) | K1,K3 | rules-2024.2 | accepted |
Officer corrections (training data queue)
- No corrections yet.
How results are explained
- • Entity matches: weighted reasons, conflicts and score (Network tab)
- • Legal suggestions: element-by-element evidence matrix (Legal tab)
- • Risk scores: contributing factors, uncertainty, validation (Risk tab)
- • AI answers: passage-level citations to immutable originals