Skip to content

The cooperative

The structure that makes everything else possible.

Africa's Voice is structured as a cooperative — a legal form registered in Nigeria under the Co-operative Societies Act. The cooperative owns the data. Members own the cooperative. This is the architecture, not a feature.

01

What the cooperative does

The Africa's Voice Data Cooperative is the legal entity that holds rights to all contributor data. It collects member voice recordings, manages consent, distributes dividends from data sales, and represents members in negotiations with the operating company. It is governed by an elected member council with representation from each language community.

02

Who members are

Any Nigerian adult who records voice data through the Olu Earn app and reaches 20 accepted contributions becomes a full cooperative member. Members receive a written certificate, a member ID, and full voting rights at the cooperative's annual general meeting. Membership is free and cannot be lost without due process.

03

How dividends work

Forty percent of all revenue from dataset sales and API usage flows to a contributor dividend pool. The pool is distributed quarterly, with each member's share calculated from their accepted contribution count for the quarter. Top-tier contributors who consistently produce high-quality data receive bonus multipliers. The exact formula is published on the member dashboard.

04

Governance and oversight

The cooperative is governed by a five-member elected council, one elected per language community (Igbo, Yoruba, Hausa, Pidgin, Nigerian English). An Independent Oversight Officer audits the relationship between the cooperative and the operating company quarterly. The cooperative can terminate its data licensing agreement with the operating company on six months' notice if member rights are violated.

Consent architecture

The consent architecture

Every audio file enters the corpus with one of three consent tiers attached:

Tier 1

Research

May be used by academic researchers and non-commercial AI development. Cannot be used to train commercial products. Default tier — every contribution starts here.

Tier 2

Commercial

May be used to train AI systems that are sold commercially. Contributor opts in explicitly with informed consent. Higher payout tier.

Tier 3

Public domain

May be used freely under a Creative Commons licence. Contributor opts in. Cannot be revoked. Used for academic benchmarks and open science.

Every consent decision is logged immutably. Withdrawal requests are honoured within 30 days. The full consent chain is queryable per file by enterprise buyers performing AI ethics due diligence.

vs. existing alternatives

How this is different

Mozilla Common Voice
Read-aloud only. No domain coverage. No paired transcripts. Public domain only — no commercial tier. No contributor ownership.
DSN African Voices (Gates Foundation)
Research-tier dataset. No API. No commercial consent chain. No cooperative ownership. Excellent academic resource — not built for enterprise use.
Whisper / Google Speech / AWS
Trained on whatever data the company collected — provenance largely undocumented. African languages either fail entirely or perform far below usable thresholds.