TUNIS — The white paper on artificial intelligence produced by ATUGE, the association of Tunisian graduates of French grandes écoles, sets out a national strategy to 2030 built on an uncomfortable pairing of numbers: Tunisia has 4,120 developers per million inhabitants — one of the highest densities in North Africa — and an AI adoption rate of 12.7 percent at the end of 2025.

The document, titled TunisIA — Building our future in the era of AI, was presented on 21 July in Tunis during the third edition of the Tunisia Global Forum and its full content was set out this week. It was drawn up by a 22-member committee working voluntarily over several months, and contains 41 recommendations organised into three axes and two levers. It is published open source.

"This white paper is meant to be shared, discussed and enriched at the pace of technological change," its authors write, inviting public, private and non-profit actors to contribute.

Amine Aloulou, one of those behind the initiative, has described the document not as a fixed programme but as a starting point for collective work.

The diagnosis

The authors' framing is that AI has stopped being a technology choice and become basic infrastructure that redefines the competitiveness of states — a general-purpose technology diffusing across every sector and reorganising value chains.

For Tunisia they identify a double exposure. On the upside: a large pool of trained engineers. On the downside, four constraints they name explicitly — the absence of large-scale data centres, the lack of sovereign compute capacity, brain drain, and the under-equipment of small and medium enterprises, which they note make up 90 percent of the country's economic fabric.

They single out one sector as first in the line of fire: outsourced service functions, the offshoring activity that employs a significant share of Tunisia's graduate workforce, and which generative and agentic AI transforms before it transforms anything else. That is the report's sharpest point. The part of the Tunisian economy built on exporting skilled labour at a discount is the part that automation reaches first.

Frugality as a strategy, not a constraint

The vision stated for 2030 is an economy capable of creating, adopting and deploying trustworthy AI. What distinguishes the document from most national AI strategies is its explicit rejection of imitation spending.

Given the scarcity of resources, the authors argue, Tunisia can only distinguish itself through the relevance and efficiency of its approach: a "smart and frugal" AI policy that favours upgrading what exists over building from scratch, pooling over duplication, and evaluated experimentation over hasty generalisation — judging every initiative against a single yardstick, measurable value creation and competitiveness.

Their five guiding principles follow from that. Start from real uses and needs, and size investment — particularly in compute and infrastructure — by actual demand rather than by mimicry. Build strategic capacity in stages, removing the regulatory and organisational obstacles whose lifting produces quick structural effects. Design trustworthy AI around citizens first, with data protection, cybersecurity, algorithmic transparency and risk management in a Tunisian framework aligned with, but not copied from, international standards. Retain existing talent, retrain workers in exposed occupations, mobilise the diaspora, and train decision-makers, civil servants and teachers. And organise cooperation between institutions, research, startups, established firms, civil society and international partners.

The part the white paper cannot supply

A voluntary association's document, however well constructed, is not a public policy. The specialist site Tunisie Haut Débit published an article this year under the title "National AI strategy in Tunisia: announced since 2018, still nowhere to be found in 2026" — a reminder that the gap in this file has rarely been a shortage of plans.

That is also why the report's open-source publication is more than a formality. By putting the text in public and inviting amendment, ATUGE has made it possible for the document to be adopted, criticised or ignored in the open rather than circulating between institutions. Whether any of the 41 recommendations acquire an owner, a budget line and a deadline is the only test that will distinguish this white paper from the ones that came before it.