Stakeholder Impact Assessment

Identify who is harmed by each bias pattern found, assess severity and urgency, then apply the relevant regulatory framework.

🤖 Kai — Auditor Advisory Note

Bias in AI is not abstract — it has real consequences for real people. The purpose of this assessment is to move from "we found a gap" to "here is who is hurt, how badly, and what the law says about it." Regulators and courts will ask exactly these questions.

Applicable Regulatory Frameworks — Reference Before Completing the Matrix

🇪🇺

EU AI Act (2024)

Classifies AI systems by risk level. Employment, education, and law enforcement AI are "high-risk" and face mandatory bias testing, transparency reports, and human oversight requirements before deployment.

Applies to my system? ⬜ Yes — High Risk ⬜ No

🇪🇺

GDPR — Article 22

Individuals have the right not to be subject to purely automated decisions that significantly affect them — including job screening, loan approval, or exam grading. They must be able to request human review.

Applies to my system? ⬜ Yes ⬜ No

🇺🇸

EEOC Guidance (US)

AI hiring tools are subject to anti-discrimination law under Title VII. If a tool has an "adverse impact" on protected groups (race, sex, religion, national origin), employers can face legal liability — even if discrimination was unintentional.

Applies to my system? ⬜ Yes ⬜ No

Stakeholder Impact Matrix — Complete for Each Bias Finding

#	Harm Severity	Urgency (1–5)
1	Individual harm Systemic harm Structural harm	LowHigh
2	Individual harm Systemic harm Structural harm	LowHigh
3	Individual harm Systemic harm Structural harm	LowHigh

Prioritised Remediation Actions — Rank Your Top 3 Recommendations

Priority 1 — Immediate

Address before deployment / continued use

Priority 2 — Short-term

Address within 90 days of audit

Priority 3 — Ongoing

Embed in regular audit cycle