Mauritius's National AI Strategy and the FAIR Guidelines

On 9 April 2026 the Government of Mauritius launched its National Artificial Intelligence Strategy together with a set of guidelines known as FAIR, developed with support from the United Nations Development Programme. Two months later the 2026/27 Budget named Leveraging AI and Digitisation as the first of its seven strategic pillars.

Taken together these are the clearest signal yet that AI in Mauritius has moved from something individual firms experiment with to something the state has a stated position on. This is a plain-language read of what changed and what it means if you are deploying AI here.

What the strategy actually covers

The strategy sets out a roadmap for integrating AI across sectors including manufacturing, healthcare, agriculture, fintech, transport and citizen services. Its framing is explicitly economic: it names falling productivity, an ageing and declining population and moderate growth as the national problems AI is being asked to help with.

That framing matters for anyone selling or deploying AI locally. The state's interest is in productivity and public service delivery, not in AI as a technology in itself. Proposals that speak to measurable output tend to land better than proposals that lead with capability.

FAIR: Fairness, Accountability, Inclusiveness and Responsibility

FAIR is the governance half. It sets expectations for how AI systems are designed, deployed and managed, and it is framed around the whole lifecycle rather than the moment of deployment: design, development, deployment, monitoring, and eventual decommissioning.

Two things about that lifecycle framing are worth absorbing:

• Monitoring is treated as part of the system, not as maintenance. A model that was accurate at launch and has since drifted is not a well-governed system that needs attention; it is a system that is currently out of compliance with its own design intent.
• Decommissioning is named explicitly. Very few organisations plan for how an AI system is retired, what happens to the data it accumulated, and how decisions it made remain auditable afterwards. Naming it in the guidelines makes it a question you should expect to be asked.

What this changes in practice

For most private-sector organisations, the strategy does not impose a new licensing regime or a filing obligation. What it does is establish a reference framework that a regulator, a client or a counterparty can point at. That has a practical consequence: we did not think about it becomes a materially weaker answer than it was in 2025.

If you are in financial services, the picture is more concrete, because the Financial Services Commission has its own AI expectations that sit alongside the national framework. We cover those separately in what the FSC's AI rules actually require.

If you process personal data with AI, the binding instrument is not the strategy but the Data Protection Act 2017, which already requires impact assessments for high-risk processing including automated decision-making. That is covered in the Data Protection Act and AI.

The money, briefly

The Budget 2026/27, tabled on 19 June 2026 under the theme Future Ready Economy, attached concrete measures to the strategy. The headline is a commitment to train or enable 50,000 people in AI skills over the year, and the breakdown tells you more than the total does:

• 25,000 Mauritians in developer, entrepreneur and SME-owner roles
• 12,000 Grade 9 students receiving AI-powered learning tools
• 8,000 secondary school teachers
• 5,000 public officers

Alongside the skills programme: a ten-year income tax holiday for qualifying startups; for manufacturing, an investment tax credit of 15% annually over three years, totalling 45% through 30 June 2029; and a new high-tech Special Economic Zone at Cote d'Or offering 100% foreign ownership, a special electricity tariff for data centres, VAT recovery and fast-track work permits for specialised foreign staff. A Golden Visa scheme covers fintech, AI, biotechnology, renewable energy and treasury investments.

This built on the previous year's Budget 2025/26, which established an AI Innovation Start-Up Programme under the Ministry of Information Technology, a dedicated AI Unit at MITCI, Rs 25 million to equip ministries with AI tools, and tax deductions on AI investments up to Rs 150,000 for start-ups and MSMEs.

We would treat the skills commitment as the most consequential of these for most firms. The constraint on AI adoption in Mauritius has rarely been the software; it has been finding people who can specify a problem well enough for a system to solve it, and who can tell when the answer is wrong. Note also that 5,000 of those places are for public officers, which is a demand signal for anyone selling into government.

What we would do about it

1. Write down where AI already sits in your organisation. Most firms are further along than they think, informally, through staff using tools nobody approved.
2. Map each use against the FAIR lifecycle. The gap is almost always in monitoring: someone can tell you how a system was built and nobody can tell you how it is performing now.
3. Decide who owns the answer. The governance question that matters is not which framework you follow but which named person is accountable when an output is wrong.

None of this requires a large programme. It requires a decision, written down, that you can show someone.

Common questions

Is the National AI Strategy legally binding?

The strategy and the FAIR guidelines are a policy framework rather than a statute. They do not create a licensing regime on their own. The binding obligations that apply to most AI deployments come from existing law, principally the Data Protection Act 2017, and from sector regulators such as the Financial Services Commission. The strategy matters because it establishes a reference point that regulators, clients and counterparties can hold you to.

What does FAIR stand for?

Fairness, Accountability, Inclusiveness and Responsibility. It is framed around the full AI lifecycle: design, development, deployment, monitoring and decommissioning, rather than only the point at which a system goes live.

Does this apply to a small firm, or only to large organisations?

The framework is not scoped by headcount. In practice a small firm's obligations are lighter because its processing is usually simpler, but the questions are the same: what does the system do, who is accountable for its output, and how would you know if it stopped working properly.

When was the strategy launched?

The National AI Strategy and the FAIR Guidelines were launched on 9 April 2026, developed with support from the United Nations Development Programme. AI was subsequently named the first of seven strategic pillars in the 2026/27 Budget.

What incentives are available for AI investment in Mauritius?

The Budget 2026/27 set out a ten-year income tax holiday for qualifying startups; an investment tax credit for manufacturing of 15% annually over three years, totalling 45% through 30 June 2029; and a high-tech Special Economic Zone at Cote d'Or offering 100% foreign ownership, a special electricity tariff for data centres, VAT recovery and fast-track work permits. A Golden Visa scheme covers fintech, AI, biotechnology, renewable energy and treasury investments. The earlier Budget 2025/26 provided tax deductions on AI investments up to Rs 150,000 for start-ups and MSMEs. Eligibility conditions attach to each, so confirm current terms with the Economic Development Board before relying on them.

How many people is the government training in AI?

The Budget 2026/27 committed to training or enabling 50,000 people over the year: 25,000 Mauritians in developer, entrepreneur and SME-owner roles, 12,000 Grade 9 students receiving AI-powered learning tools, 8,000 secondary school teachers and 5,000 public officers.

Related: Choosing a provider

Talk to the people who build it →