A project of the Seed Trust · seedtrust.net
AI that heals,
not harms.
Grounding artificial intelligence in African values, policy rigour, and human dignity — for the continent and beyond.
Africa needs its own AI conscience.
Southern Africa Responsible AI (SARAI) exists because the global conversation about artificial intelligence is happening without African voices — while African communities bear disproportionate risk from poorly designed systems.
We conduct independent audits, publish policy positions, and build frameworks that centre dignity, sovereignty, and Ubuntu as non-negotiable criteria for ethical AI deployment.
SARAI operates as a project of the Seed Trust, a KwaZulu-Natal-based organisation with more than three decades of community development and institutional integrity in Southern Africa.
“Umuntu ngumuntu ngabantu” — a person is a person through other persons. AI systems that sever this relationality are not neutral tools: they are instruments of harm.
Ubuntu Standard · SARAI Intellectual Framework
Crisis-Conflict Loop
A diagnostic model showing how AI deployments in fragile contexts amplify existing crises through feedback loops between surveillance, conflict, and displacement.
Implementation Gap
Why do nations endorse AI ethics principles — yet harm continues? SARAI exposes the structural barriers that separate declaration from practice.
Ubuntu Standard
A relational ethics framework grounded in communal African philosophy. Does this system strengthen or fracture the web of human relationships it touches?
IP Sovereignty
African languages, oral traditions, and cultural heritage must not be extracted without consent, attribution, or benefit-sharing for source communities.
The RAINBOW Framework
Seven audit domains. A complete spectrum of AI risk assessment — tailored to Southern African political, economic, and social realities.
The EVAH Initiative
Ethical Validation and Audit for Health AI — Africa’s first structured audit programme for artificial intelligence deployed in healthcare settings.
Diagnostic AI Audits
Assessing diagnostic tools used in under-resourced clinics for bias, accuracy drift, and informed consent gaps.
Community Consent Protocols
Culturally appropriate frameworks for community-level consent when AI affects group health outcomes.
Training Data Sovereignty
Investigating whether African patient data used to train global health AI returns value to source communities.
Open Audit Reports
All findings published openly — accessible to policymakers, communities, and international standards bodies.
Healthcare providers, development funders, and AI developers: SARAI is seeking partners committed to accountable health AI deployment in Africa.
Where SARAI stands
Position statements and policy submissions grounded in evidence, written for decision-makers.
IP Sovereignty and African Training Data
African languages, oral traditions, and cultural heritage are being used to train large language models without consent, attribution, or return of value. SARAI’s foundational position on what must change — and the legal and ethical conditions required.
The Crisis-Conflict Loop: Diagnostic Report
How AI surveillance tools deployed during conflict amplify displacement — and why existing humanitarian AI guidelines fail to account for African political geographies and community structures.
Implementation Gap: Southern Africa Mapping
A mapping of the structural distance between national AI ethics endorsements and measurable policy outcomes across SADC member states. What the gap costs in human terms.
Ready to act on responsible AI?
Whether you’re a policymaker, funder, researcher, or technology deployer — SARAI offers audit services, policy consultation, and partnership on the EVAH Initiative.
South Africa
seedtrust.net
Policy · Media
SARAI is a project of the Seed Trust ·
Howick, KwaZulu-Natal · South Africa ·
© 2026 Seed Trust. All rights reserved.