As investment in Arabic AI grows across the Middle East and beyond, the development of effective models remains complex and demanding, according to experts interviewed by AGBI.
Artificial intelligence spending in the Middle East, Turkey, and Africa is projected to hit $5 billion this year, with a compound annual growth rate of 35%, reaching nearly $12 billion by 2028, according to research by global tech market analysts IDC. Currently, 60% of organizations in the Middle East have identified AI as their primary focus area for investment.
Despite this momentum, building large language models (LLMs) for Arabic presents unique hurdles. Vasudha Khandeparkar from Grant Thornton UAE highlights the intricate sentence structures, numerous dialects, and rich but under-digitized linguistic history of Arabic as major obstacles to AI model training and performance.
A study by researchers at the University of Jordan, published in Springer Nature’s BMC Infectious Diseases journal, compared how well AI models performed on English versus Arabic AI queries related to infectious diseases. The results showed consistently stronger outcomes in English, underscoring the performance gap that still exists.
However, there are emerging breakthroughs. Canadian AI firm Cohere recently unveiled a multilingual model—Command R7B Arabic—that is gaining attention for its strong capabilities. It has been designed to tackle Arabic-specific issues like complex morphology and regional dialects, excelling in tasks such as document summarization, question-answering, and task automation.
Khandeparkar noted that scaling such models in the region adds another layer of complexity due to the technical architecture and infrastructure needed. “This makes the R7B model particularly promising for businesses operating in Arabic-speaking markets,” she said.
But the barriers go beyond language and infrastructure. Melih Murat, Associate Research Director at IDC, pointed out that many organizations still lack the foundational strategies needed to integrate AI effectively. Challenges include limited access to quality data, inadequate internal policies for AI governance, and insufficient employee readiness.
Murat warned that unless these structural issues are addressed, companies may find it difficult to extract real business value from foundation models and advanced AI tools.
