Arabic Artificial Intelligence: Do We Make or Consume? Sovereignty and Arabic Language Models

We are a nation whose hundreds of millions speak one language, yet the most powerful models are trained on data dominated by English, so Arabic arrives in them with a weaker share, like a guest at a table that is not its own (Joshi et al., 2020). Is our fate to import even the way the machine thinks about our words? Or does this moment, while the doors of the race have not yet closed, offer a chance to be makers rather than consumers?

A question of civilisation, not of luxury

Civilisations are measured by their ability to produce the tools of their age, not to acquire them. The scholars of Islamic civilisation, who transmitted the sciences of other nations from the eighth century CE, did not stop at translation: they added algebra with Al-Khwarizmi in the ninth century and optics with Ibn al-Haytham in the eleventh. The word “algorithm”, on which AI systems rest, derives through Latin from the name of Muhammad ibn Musa al-Khwarizmi. Yet here the heir stands as a recipient of others’ algorithms, paying for use and exporting raw data to re-import it processed.

Stanford’s AI Index (2025) notes that the United States produced 40 notable models in 2024, against 15 for China and 3 for Europe combined. Nick Couldry and Ulises Mejias, in their 2019 book, describe the extraction of data from human lives for the benefit of a few companies as “data colonialism”, a description that remains debated; others see its effects as harsher on countries of the Global South.

A view that departs from the familiar

First: the rhetoric of self-flagellation, which sees Arabs as perpetual consumers, is an oversimplification that history does not support; technology has no fixed destiny. In the 1960s South Korea was at an income level similar to or lower than that of many Arab countries, then became a major technology maker within a generation, through will, planning and staying power.

Second: the “make or consume” dichotomy may itself be a mental trap. No one makes everything; even the giants specialise in links of a global chain. The better question is: what do we make, and where do we place our weight? Where we hold an advantage: the richness of Arabic in morphology and meaning, knowledge of cultural, religious and legal contexts that an imported model does not grasp, and sectoral data in energy, pilgrimage, trade and health.

Language is the heart of the battle

Language is a structure through which thought and values take shape. When a model is trained on texts in which Arabic is scarcely visible, we receive a lens tinted by another culture, and a community’s absence from the training data weakens the machine’s representation of it (see: “Algorithmic Bias”). So building authentic Arabic models is not merely a technical project but knowledge sovereignty, and a nation’s right to have the machine that serves it think in its own tongue.

Resources: computing, talent and data

Three elements govern the capacity to produce: computing, talent and data. The region, with the Gulf at its heart, has the ingredients for computing, namely energy and capital, though reliance on imported chips remains. But talent and data cannot be bought with money alone; they are cultivated by the patience of education, a culture of research and freedom to experiment. The gravest trap is to confuse owning the infrastructure with owning the knowledge that drives it, building data centres to run other people’s models. Many believe progress accelerates by accumulation, though researchers do not agree that this will continue; even so, waiting is not a neutral stance but a loss, and speed alone is not enough unless joined to the right direction.

Five practical pillars

1. Data sovereignty: Arabic data is a strategic wealth, to be preserved and invested within the region.
2. Investing in the mind before the stone: research universities, and retaining talent.
3. Smart specialisation: links of the value chain where our language and context give us the edge.
4. Arab integration: no single Arab country holds the market, talent and capital, but together the Arabs hold the ingredients.
5. An ethical framework rooted in our values (see: “Responsible Artificial Intelligence”).

The National Strategy for Data and AI (2020) connects to Saudi Vision 2030 in building a digital economy and moving from rent to knowledge. The measure of success is not the number of centres opened, but their ability to nurture an Arab researcher who, from Riyadh, Cairo or Tunis, creates an algorithm the world benefits from as it benefited from Al-Khwarizmi’s.

To make with our spirit

Among the most useful things Arabs can offer in this century, beside models, is an ethical vision of technology that proceeds from the principle that the human being is honoured and not to be used as a tool, and that knowledge is a trust, not an instrument of domination. So we make not merely to catch up with the caravan, but to add to humanity’s journey a mark that bears our spirit.

This century may well divide nations into those who build the machine’s mind and those to whom their machine is dictated. The Arabs, with a living language, a scientific history, resources and youth connected to the digital world, stand at a rare crossroads. The answer is not yet settled; we pursue it with today’s decisions, relying on God and taking the means: curricula we teach, researchers we retain, data we protect, and a language in which we insist the machine should think. We ask God to grant us success in what befits our nation; and the future, God willing, is written by hands that work.