How the roll-up gate works

How the AI Impact Assessment Builder works

This tool grades each identified impact of one AI system on severity, likelihood, reversibility, and affected population, reduces it by its mitigation to a residual, then rolls the assessment up to its worst unmitigated impact. It reads 4 impacts and rolls them into one verdict — the worst, never the average — then a gate that can only make it worse.

The 4 impacts

01I-01 · wrongful benefit denial to applicantsfix firstgate · severe + irreversible + whole population, mitigation only partial · residual 40UNACCEPTABLE

02I-02 · biased ranking disadvantages a groupresidual 29 · mitigated in place, not eliminatedRESIDUAL

03I-03 · privacy exposure of sensitive dataresidual 11 · mitigation in place with oversightMITIGATED

04I-04 · over-reliance erodes staff judgmentresidual 37 · thin mitigationRESIDUAL

DEFENDED

the worst unmitigated impact is MITIGATED - every residual is brought low

GAPS REMAIN

the worst impact is a RESIDUAL that's controlled but not eliminated, and nothing is unacceptable

EVIDENCE GAPS PRESENT AS DESCRIBED

any impact is UNACCEPTABLE - a residual over the line, or a severe, irreversible harm to a vulnerable group that isn't fully mitigated

Irreversible-harm-to-vulnerable-group gate · worsen-only

An impact that is severe AND irreversible AND on a large or vulnerable population is forced to UNACCEPTABLE unless it is fully mitigated with human oversight — even if its residual score looks moderate. Any UNACCEPTABLE impact makes the whole assessment read EVIDENCE GAPS PRESENT AS DESCRIBED, because an assessment is only as defensible as its single worst unmitigated impact.

EVIDENCE GAPS PRESENT AS DESCRIBED

Why: One impact is well-controlled (residual 11) and two sit at moderate residuals (29, 37), so the mean reads 29 - but the assessment takes the worst, not the average. I-01, a wrongful benefit denial, is a severe, irreversible harm to a large population and its mitigation is only partial: the gate forces it to UNACCEPTABLE even though its residual computes to 40. Any unacceptable impact makes the assessment read EVIDENCE GAPS PRESENT AS DESCRIBED. 1 UNACCEPTABLE (gated) · 2 RESIDUAL · 1 MITIGATED · worst residual 40 · mean 29 · worst-not-average.

Fix firstI-01 - wrongful benefit denial to applicants. It's a severe, irreversible harm to a vulnerable group that isn't fully mitigated, which the gate treats as UNACCEPTABLE on its own. Bring its mitigation to full - in place with human oversight - first; that's the one impact holding the assessment at EVIDENCE GAPS PRESENT AS DESCRIBED.

No single impact is the crisis — the clustering is. The roll-up takes the worst, the gate escalates when trouble compounds, and where something is short, the fix first row names it. A working aid, not legal advice. This grades the impacts you enter from your own ratings — it is not the assessment of record, a fundamental-rights determination, a certification, or an audit. It does not discover impacts you didn't list, decide what residual is lawful for your context, or score any person. Confirm your impact criteria and any FRIA / impact-assessment obligation with a qualified auditor or counsel.

RedHub AIthe AI that tells you when to doubt it redhub.ai/systems/ai-impact-assessment-builder