The AI sovereignty debate sounds like a policy conversation for governments and data scientists. It is not. For a creator in Cairo filming a tutorial in Egyptian Arabic, or a podcaster in Casablanca recording in Darija, the question of who trains the models and on what data determines whether their content reaches an audience or gets flattened into something generic.
Africa’s four biggest tech economies have already confronted this. South Africa, Kenya, Nigeria, and Egypt have each drafted AI strategies that admit dependence on Google, Microsoft, Nvidia, and Meta for infrastructure, as Ananya Bhattacharya reported for Rest of World. That is not a confession of failure. It is a realistic starting point. Rachel Adams, founder of the Global Center on AI Governance, told Rest of World that Africa’s push for digital sovereignty cannot mean total independence from global AI supply chains. It can mean stronger control over sensitive data, better public procurement rules, investment in local infrastructure and skills, African language data sets, and clearer accountability for foreign AI providers. The numbers underline the gap: according to the World Economic Forum, Africans comprise 18% of the world’s population but the continent has less than 1% of global data center capacity.
MENA creators should read this as a cautionary tale, not a distant one. The same structural dependence applies across the Arab world. When a creator relies on an AI tool trained on English-language content to generate captions, thumbnails, or scripts, they are outsourcing the cultural and linguistic framing of their work to a system that does not know their audience.
The Pope’s encyclical and the missing Muslim voice
Pope Leo XIV released his first encyclical, titled Magnifica Humanitas, on May 29, 2026, warning that AI power should not be concentrated in the hands of a few private companies. The timing is relevant. The same week, Rina Chandran reported for Rest of World that Egypt banned the use of AI to interpret the Quran in 2026 because of concerns that chatbots were moving Muslim users toward Western values. Two signals from different institutions pointing at the same problem: the people building AI do not represent the people using it.
Brian Patrick Green, director of technology ethics at the Markkula Center for Applied Ethics at Santa Clara University, said that AI should be able to serve everyone, and that means it needs to know about the religions of the world. The gap is not just religious. It is cultural, linguistic, and commercial. When a platform’s AI moderation tool flags a perfectly normal Arabic phrase as inappropriate, or when a recommendation algorithm trained on English-language content fails to surface Arab creators, the result is lost reach and lost revenue. Egypt’s ban is a state-level response to a problem that creators experience daily at the algorithmic level.
The ex-coded creator
Benjamin Rosman, a professor of computer science at the University of the Witwatersrand, highlighted at a Vatican conference that over 2,000 languages are spoken across Africa, many of which are low-resource languages with insufficient data to train AI systems effectively. Arabic is not a single language. It is a spectrum of dialects that vary across the Gulf, the Levant, North Africa, and the diaspora. A model trained on Modern Standard Arabic and English will not understand a creator speaking in Tunisian Arabic or Iraqi dialect.
Joy Buolamwini, AI researcher and founder of the Algorithmic Justice League, warned at the same Vatican conference about the rise of the ex-coded individual: someone who is harmed, discriminated against, or exploited through AI systems. For an Arabic-language creator, being ex-coded means the algorithm does not push your content because it cannot parse your language. It means the auto-caption tool mangles your words. It means the brand-deal matching platform trained on English-language creators does not have a category for you. That is not a future risk. It is the current state of play.
For an Arabic-language creator, being ex-coded means the algorithm does not push your content because it cannot parse your language.
What creators and platforms can actually do
The sovereignty framework Adams described for Africa translates directly to the MENA creator economy. Stronger control over data means creators and platforms demanding that AI companies disclose what data was used to train the models that power their tools. Better public procurement rules means regional platforms like Anghami and Shahid insisting that the AI services they license include Arabic-language training data. Investment in local infrastructure means governments and private capital funding data centers and annotation projects that build Arabic-language datasets.
The goal is not total independence. It is strategic leverage. A creator who understands that their content is training data for someone else’s model can make different choices about which platforms to prioritize and which AI tools to trust. A platform that invests in Arabic-language AI training is building a moat that English-first competitors cannot easily cross.
The Pope’s encyclical, Egypt’s ban, and Africa’s infrastructure gap are not separate stories. They are the same story told from different angles. The question is whether MENA creators will wait for someone else to build the models that represent them, or whether they will start demanding the data sovereignty that makes those models possible.