01 — Overview
BITC is a framework for a trust infrastructure aimed at African public institutions and regulators. It seeks to make the AI systems used in sensitive areas (health, finance, predictive justice, administration) inspectable, documentable and monitorable.
- Field
- AI Governance & Risk
- Alignment
- African Union AI Strategy
- Key component
- AI digital passport
- Stage
- Prototype
02 — Problem
Most AI models used in Africa are developed outside the continent, trained on exogenous data and rarely assessed against criteria adapted to African linguistic, cultural and legal realities.
Without a technical audit framework, administrations are exposed to algorithmic bias, to unregulated export of sensitive data and to a loss of decision-making control over systems that directly affect users.
03 — Objective
To design a technical and methodological foundation to: document each deployed AI system (data origin, purpose, limits, accountability); assess statistical disparities and bias risks; monitor post-deployment performance and detect drift; and produce verifiable evidence of compliance with African and international frameworks.
04 — Proposed solution
- Registry & AI passport: a documented identity record for each model, including training-data traceability.
- Risk & bias engine: quantitative assessment of statistical parity and robustness.
- Continuous audit: post-deployment monitoring (data drift, performance degradation).
- Compliance gateway: risk classification and generation of a trust score.
05 — My role
Conception of the framework, drafting of specifications (AI passport schema, audit engine logic) and development of the audit prototype. This project is a personal applied-research effort and is not part of any institutional commission at this stage.
06 — Technology / methodology
Schema specification in JSON Schema, a Python prototype for calculating the disparate impact ratio (DIR) and the trust score, aligned with risk classifications (NIST AI RMF, logic comparable to the EU AI Act).
07 — Governance & policy dimension
The framework is built around existing references:
- African Union Continental AI Strategy (2024) — local capacity, African languages, national data protection.
- UNESCO Recommendation on the Ethics of AI (2021) — transparency, explainability, fairness.
- Malabo Convention — sovereignty of cross-border flows and personal data protection.
08 — Current status
Prototype. The conceptual framework, the AI passport schema and a bias-audit prototype exist. No production deployment has taken place. No institutional partner is engaged to date.
09 — Evidence
- AI digital passport schema specification (JSON Schema).
- Bias-audit and trust-score prototype (Python).
- Associated analysis note (Policy Brief #01).
10 — Roadmap
- Consolidate the AI passport specification and document the data model.
- Extend the audit prototype and add reproducible tests.
- Study the feasibility of a supervised pilot with an academic or regulatory partner — to be defined.
Any pilot or deployment step depends on an institutional partnership that is not yet established.
11 — Potential impact
In time, such a framework could help regulators and administrations assess the AI systems they use, document automated decisions and strengthen public trust — subject to validation by competent institutions.
12 — Documents
13 — Links
- Public code repository
- Related projectsAfrican Digital Sovereignty Observatory