Community Constitutions are the heart of Orchestrate.Agency’s approach: human-readable, negotiable documents that translate a community’s laws and values into specific, measurable rules an AI system must follow. The four below are illustrative drafts: Grounded in the real laws and AI systems of each jurisdiction - Kenya, Canada, Japan and the UAE. While each is detailed enough to be enforced, they are intended as starting points for the regulators, civil society, and affected communities who would negotiate, amend, and adopt the versions that govern real systems. Each defines what data may be used, what disparities are unacceptable, what decisions require human review, and how appeals work.
Note: These drafts are open resources, intended to be adapted by individual jurisdictions.
This illustrative draft is grounded in the Data Protection Act, 2019 and CBK Digital Credit Providers regulations. Sets fairness rules for AI lending decisions: approval-rate disparities above 5% between demographic groups trigger automatic human review, every denial must carry a plain-language “but-for” explanation, and mobile-money history is recognised as a valid creditworthiness signal for thin-file borrowers. Prohibits using a borrower’s handset model as a proxy variable and bars seasonal agricultural income from being treated, by itself, as elevated risk. View the full constitution.
This illustrative draft is based on Bill 149 (Working for Workers Four Act), AIDA, and the Canadian Human Rights Act
It governs AI hiring and screening tools: assessment accuracy must be equal across demographic groups, advancement-rate disparities above 5% pause screening, and international credentials and career gaps may not be penalised. Strictly prohibits facial recognition, emotion detection, accent analysis, and “culture fit” or “leadership potential” scoring in video interviews and predictive analytics and most importantly with a proxy-variable test applied to every input.
This illustrative draft is grounded in the Pharmaceutical Affairs Law/PMDA, the APPI, and the Hiroshima AI Process principles. It
governs AI diagnostic support (endoscopy, screening, and preventive-health systems) with particular care for Japan’s elderly population: every recommendation is advisory, the attending physician retains full authority, and any patient may request a human-only assessment. Diagnostic accuracy is tracked across four age cohorts, a disparity above 5% points forces investigation, and systematically lower accuracy for patients 75+ pauses the system - a threshold no vendor configuration can change.
This illustrative draft is based on Federal Decree-Law No. 33 of 2021, the UAE AI Ethics Charter, and DIFC Law No. 1 of 2025. The constitution
governs the MoHRE “Eye” document-verification system across 13 permit types and 80+ nationalities: it may verify document authenticity and eligibility but is barred from judging “suitability,” institutional prestige, or integration likelihood. Approval-rate disparities above 5% between nationalities pause processing; recommendations require 95% confidence to approve and 98% to deny; and facial recognition or ethnicity estimation from photos is strictly prohibited.
Copyright © 2026 Orchestrate - All Rights Reserved.