INDEPENDENT INITIATIVE · DOI 10.5281/zenodo.24043894

The Global Reference for Algorithmic Biases

A rigorous scientific knowledge base to address global concerns about AI biases alNGOsside other institutions.

250
Documented biases
1,511
Global regulatory links
883
Protected populations
409
Scientific sources

Addressing global concerns

Leaders of major AI laboratories, researchers, the Pope, NGOs, and international institutions have been warning for several years about the dangers of current developments in artificial intelligence.

This initiative is a citizen response to these warnings: A rigorous scientific knowledge base to address global concerns about AI biases alNGOsside other institutions.

Each bias is classified according to the Taxonomy TriAxis (Origin × Lifecycle × Specific) and linked to applicable Regulations.

O

Origin Axis

Where does the bias come from? Systemic (O1), computational (O2) or human-cognitive (O3).

L

Lifecycle Axis

At what stage? From L1_historical (collection) to L6_deployment (production).

S

Specific Axis

Specific nature: proxy, feedback, sampling, annotation... (30 categories).

Framework de Governance

AEXIA - Governance de l'Explainable AI

AEXIA (Accountable, Ethical, eXplainable, Interpretable, Auditable AI) is an integrated framework that articulates the methodological dimension of explainability, organizational maturity, and the individual maturity of deployed systems.

Decouvrir AEXIA
🧭

Dual axis

AEXIA-O (organization) and AEXIA-S (system), crossed into a decision matrix.

🧩

DCME, ECS, RAD

Three original contributions: multi-scale causal explanation, consistency score, responsibility debt.

🛡️

4 mecanismes

Weakest link, evidentiary gates, AI-specific downgrading, risk proportionality.

Cite this work

If you use this API or this data in your scientific work, please cite:

@software{sidibe_2026_ia_biais,
  author    = {SIDIBE, Mohamadou},
  title     = {AI BIAIS Initiative - API REST},
  year      = 2026,
  publisher = {Zenodo},
  version   = {1.0.0},
  doi       = {10.5281/zenodo.24043894}
}

Join the initiative

Researchers, policymakers, NGOs, institutions - this database is open. Contribute, cite, criticize. Together, let's make AI a fair tool for everyone.