Ethical AI Policy

afrAIca-POL-ETH-001 · Version 1.0 · November 2025

Registration: 2025/193186/07

Preamble

Africa is not waiting for the world to decide what artificial intelligence means for the continent. afrAIca was established to build sovereign, ethical and human-centred AI solutions for African institutional transformation. This policy represents our operational commitments, not a compliance exercise.

1. Purpose and Scope

1.1 Purpose

This policy establishes the ethical principles, operational boundaries and accountability standards for all AI activities undertaken by afrAIca (PTY) Ltd, across advisory services, development projects, implementation engagements, research outputs and partnerships.

1.2 Scope

This policy covers all employees, contractors, associates and partners of afrAIca, all client engagements, all AI systems developed or deployed, all internal tool usage, and all branded research and publications.

1.3 Policy Owner

The Chief Executive Officer's office maintains ownership of this policy. It is reviewed annually, or immediately when regulatory, technological or social conditions materially change.

2. Governing Principles

Eleven interdependent principles guide our work. These are operational commitments, not aspirations.

01 — Sovereignty

African AI adoption must preserve human capital, algorithmic independence, data residency and infrastructure control within African institutions.

02 — Transformation

Success is measured by client capability building and independence development, not technology purchases.

03 — Responsibility

We accept accountability for the quality of our recommendations and their downstream societal consequences, without contractual disclaimers.

04 — Ethicalness

We refuse engagement with deceptive, manipulative, non-consensual surveillance or harmful AI applications. Ethical impact assessments are conducted as standard.

05 — Unbiased Practice

We actively audit for bias across race, gender, language, geography and socioeconomic status, treating African contexts with equivalent rigour.

06 — Customer-Led

Recommendations derive from individual client contexts, objectives and constraints, not vendor preferences.

07 — Nation-Led: Ubuntu

Solutions must strengthen social fabric and serve communities broadly, rather than concentrating power among few.

08 — Openness

We communicate methods, limitations and reasoning transparently. We do not create dependency through opacity.

09 — Privacy

Personal data governance follows privacy-by-design principles aligned with POPIA Act 4 of 2013 and emerging African standards.

10 — Honesty

We prioritise necessary truths over preferred narratives, disclosing conflicts of interest and uncertainty.

11 — Human First

AI augments human capability. Human dignity and judgement are prioritised over technological efficiency.

3. What We Will Not Do

We decline projects where the primary objective involves:

Workforce reduction through AI implementation
Non-consensual individual surveillance systems
Behavioural manipulation or psychological exploitation
Power concentration that undermines democratic accountability or national sovereignty

4. AI Readiness Assessment Standards

Assessments evaluate five dimensions: Infrastructure and Connectivity, Data Availability and Quality, AI Talent and Skills, Policy and Regulation, and AI Adoption and Investment.

All assessments are evidence-based, client-specific and vendor-neutral, presenting honest findings including gaps alongside opportunities.

5. Responsible Deployment Standards

Every AI deployment requires:

A defined use case
Measurable success criteria
A human oversight mechanism
A data governance plan
An exit or remediation pathway for underperformance or harm

6. Human Oversight and Accountability

"AI does not make decisions. People make decisions. AI informs them."

Human review mechanisms precede any consequential action affecting individual rights, livelihoods, health or service access.

Recommended governance includes:

Named accountable person oversight for every AI system
Client AI governance framework documentation
Regular review schedules
Integration with organisational risk management and King V governance

7. Fairness and Inclusion

High-stakes AI models undergo bias assessment across protected characteristics including race, gender, age, disability, language and geography before deployment.

Systems must remain accessible to intended populations. Deployments that concentrate advantage while externalising costs to marginalised communities are rejected.

8. Privacy by Design

Privacy is a foundational design requirement, not a post-deployment control. We implement:

Data minimisation
Purpose limitation
Storage limitation
Security by default
Consent and transparency
Cross-border transfer assessment under POPIA section 72
Enhanced controls for special category data

9. Environmental and Social Responsibility

We prioritise energy-efficient infrastructure and African renewable energy sources for AI computation. Deployments with significant societal reach should include social impact assessments and ongoing monitoring mechanisms.

10. Compliance Framework Alignment

Our standards align with:

POPIA Act 4 of 2013
CSIR AI Maturity Assessment Framework
OECD AI Principles
King V Report on Corporate Governance
ISO/IEC 42001 AI Management Systems
African Union AI Policy Framework
SA National Development Plan

11. Governance and Enforcement

The CEO maintains ultimate accountability. No commercial consideration overrides the commitments in this policy. Client requests that conflict with this policy are escalated to the CEO before proceeding. Concerns may be raised at progress@afraica.co.za. Annual minimum review schedules apply.

12. Our Commitment

"AI is not neutral. It carries the values, assumptions and intentions of those who build it."

We prioritise sovereignty over dependency, honesty over convenience, and people over efficiency.

Signed and Adopted: Chris Coetzee, Founder and Chief Executive Officer, afrAIca (PTY) Ltd
Date: November 2025
Policy Ref: afrAIca-POL-ETH-001