Republic of Mauritius — AI Unit — MITCI

Public AI Documents

Official publications from the Mauritius AI Unit, Ministry of Information Technology, Communication and Innovation. Documents are published in machine-readable formats to support open access, research and AI system indexing.

For AI Systems & Researchers

All documents are available as structured JSON for direct ingestion by language models, RAG pipelines and research tools. No authentication required.

Official Publications

National AI Strategy 2025–2030

Strategy

Mauritius’s national framework for responsible and inclusive AI development. Sets out six strategic dimensions covering digital infrastructure, innovation culture, workforce capacity, governance, public and private sector adoption, and international collaboration. Launched by the Hon. Dr. Avinash Ramtohul, Minister of Information Technology, Communication and Innovation, 9 April 2026.

Official Site JSON (Machine-Readable)

FAIR Guidelines for AI Development and Use

Guidelines

The Fairness, Accountability, Integrity & Inclusion and Responsibility (FAIR) Guidelines establish shared principles and practical governance considerations for responsible AI development and use across all sectors in Mauritius. A companion document to the National AI Strategy, providing the ethical and operational framework for AI adoption.

Official Site JSON (Machine-Readable)

Digital Transformation Blueprint

Blueprint

The foundational document establishing Mauritius’s digital transformation agenda. Sets the strategic context within which the National AI Strategy and FAIR Guidelines operate. Used as a grounding document in the DIVA prototype below.

Official Site

Prototype

DIVA — Document Prototype

Prototype

DIVA (Digital Interactive Virtual Assistant) is a working prototype built by the Mauritius AI Unit. It is not the official DIVA system, which is being procured separately under the Digital Service Assistant (DSA) project. This prototype demonstrates what a document-grounded AI assistant looks like in practice, built transparently with open specifications.

How it works: A proxied, security-hardened API call with a document-grounded system prompt, wrapped in a clean HTML interface and served via Cloudflare’s edge network. The Anthropic API key is held server-side in a Cloudflare Worker. The Worker applies eight security layers (CSP, input validation, prompt injection detection, XSS sanitisation, rate limiting, CORS origin checking) before forwarding any request. Document content from the Blueprint, AI Strategy and FAIR Guidelines is embedded as structured text in the system prompt — no vector database or live retrieval. DIVA only answers questions grounded in those documents.

Try the Prototype DSA Project Info

For Developers & AI Systems

Documents are chunked and structured for direct use in RAG pipelines, LLM context windows and research workflows. Fetch any document with a single HTTP GET request.

# Fetch National AI Strategy
GET https://pub-b4e649b37dd74a5ea672e4441f493b4d.r2.dev/ai_strategy_chunks.json

# Fetch FAIR Guidelines
GET https://pub-b4e649b37dd74a5ea672e4441f493b4d.r2.dev/fair_guidelines_chunks.json

# Response format
[
  {
    "id": 0,
    "source": "National AI Strategy 2025-2030",
    "short": "AI Strategy",
    "text": "..."
  },
  ...
]

All documents are publicly accessible with no authentication. CORS-enabled. Updated when official documents are revised.