Responsible AI Risk Matrix

Rate your product against the four Responsible AI pillars, define your mitigation strategy for each, then complete the Power Structure Analysis.

🤖 Kai — Mentor Note

The Ethics Review Board will examine your risk awareness closely. Vague mitigation strategies ("we will be careful about bias") will be challenged directly. The board expects specific mechanisms — who audits it, how often, by whom, and what triggers a suspension of the system.

Four Responsible AI Pillars — Rate Risk Level (1 = Minimal, 5 = Severe) and Define Mitigation

Pillar	Risk Rating	Mitigation Strategy — be specific
Fairness Does the system treat all groups equitably? Does it have disparate impact on protected characteristics?	1 2 3 4 5
Accountability Who is legally and ethically responsible when the system causes harm? Is there a redress mechanism?	1 2 3 4 5
Transparency Can affected parties understand how the system makes decisions? Can they contest outcomes?	1 2 3 4 5
Safety Could the system cause physical, psychological, financial, or societal harm? What are the failure modes?	1 2 3 4 5

Risk Overview — Write Your Score in Each Box

Fairness

/ 5

Accountability

/ 5

Transparency

/ 5

Safety

/ 5

Highest-risk pillar and why it is the most critical for your specific product:

Power Structure Analysis — The Three Questions Every AI Product Must Answer

Who Benefits?

List every group that gains from your AI — directly and indirectly. Include commercial beneficiaries.

Who Bears the Risk?

List groups who face harm if the system fails, is misused, or is deployed without adequate safeguards.

Who Decides?

Who controls how it is built, what data it uses, when it is updated, and when it is shut down?

Reflection: Is there a significant mismatch between who benefits and who bears the risk? If yes — what specific design changes would reduce that imbalance?