Human review
AI must not independently hire, reject, diagnose or make safeguarding findings.
ORVIA Insight is being designed for consequential use, so human authority, evidence integrity, accessibility, privacy and auditability are explicit requirements rather than footnotes.
AI must not independently hire, reject, diagnose or make safeguarding findings.
Original responses, revisions, timestamps, evidence versions and review notes should remain auditable.
The interface must not become an unintended test of disability.
Purpose, lawful basis, minimisation, retention, access controls and processor/controller clarity are required.
One organisation must never see another organisation’s participant data.
Perspective Room should not be called validated or proven before evidence supports that claim.