By: Basma Balabel

On Wednesday, April 8th 2026, the Access to Knowledge for Development Center (A2K4D) at the American University in Cairo’s Onsi Sawiris School of Business and its flagship initiative MENA Observatory on Responsible AI, co-organized a webinar titled “Large Language Models in Arab Journalism: Opportunities, Risks, and Epistemic Transformations”, in collaboration with the Arabi Facts Hub. The webinar was moderated by Osama El-Sayyad, Project Manager at Arabi Facts Hub, who was joined by three speakers; Khaled Mansour, Writer, Consultant, and Former UN Spokesperson & Human Rights Advocate, Mahmoud Hadhoud, Media Researcher & Obama Foundation Scholar at Columbia University, and Nagham El-Houssamy, Associate Director for Research, Middle East and Africa at the A2K4D and the MENA Observatory on Responsible AI.

Opening the session, Osama Al-Sayyad framed AI as a defining force in contemporary Arab journalism. Emphasizing that AI is no longer a headline in the technology section, but rather a core part of the daily workflow inside newsrooms. He backed this observation by a recent survey research conducted by the Arabi Facts Hub, involving over 249 Arab journalists. The findings reveal that more than half of the respondents use AI tools on a daily basis with over two-thirds relying on them weekly. Journalists reported primarily using AI for tasks such as research, transcription, summarization, and translation, yet paradoxically, levels of trust in these tools remain low.

Efficiency Versus Epistemic Uncertainty

Mahmoud Hadhoud explained that generative AI offers undeniable productivity gains where the minimum acceptable output is achieved very quickly, citing time savings that are difficult for journalists under pressure to ignore, yet these gains come at a cost.

Journalists report concerns over hallucinations, contextual misunderstanding, and political misrepresentation, challenges that are especially acute in a region already grappling with complex information disorder shaped by conflict, censorship, and polarization.

“This is not a question of whether we use AI” Hattab argued, “but under what conditions we can use it without sacrificing linguistic integrity or epistemic independence.”

Linguistic Inequality and Knowledge Dependency
One of the most persistent challenges discussed was the linguistic productivity gap. Large language models consistently perform better in English than in Arabic, not only in fluency, but in sourcing, contextual awareness, and verification.

This imbalance stems from structural realities since the majority of training data for large AI models is produced in English and other Global North languages. Consequently, Arab journalists often switch to English prompts to reduce errors which risks reinforcing long-term knowledge dependency.

The concern is not merely technical, but cultural. Reliance on English-language outputs can marginalize local narratives, weaken Arabic journalistic ecosystems, and entrench Northern epistemic dominance in how stories are framed and validated.

Platform Power and Embedded Values
Khaled Mansour addressed the political economy behind AI systems and platforms. Generative models, content moderation tools, and distribution platforms are largely controlled by the same handful of corporations predominantly based in Silicon Valley.

“These systems are not neutral,” Mansour emphasized. “They operate according to value frameworks shaped by profit, regulation, and geopolitical priorities.”

He highlighted how content moderation decisions, particularly around Palestine and other politically sensitive issues, demonstrate how platform governance can silence certain voices while amplifying others. The lack of sufficient investment in watermarking, detection of AI-generated media, and localized moderation further exacerbates the spread of disinformation in Arabic-speaking contexts.

Responsible AI as a Regional Dialogue
Representing the MENA Observatory on Responsible AI, Nagham El-Hassamy outlined the Observatory’s role as a bridge between regional voices and global AI governance debates. Hosted at AUC’s Access to Knowledge for Development Center, the Observatory focuses on three pillars: responsible AI for policy, for practice, and for people.

“Our work is not only about research,” she explained, “but about translating evidence into policy recommendations and ensuring that voices from the region are present in global conversations.”

Initiatives like this webinar, she argued, are essential for building collective capacity, fostering trust, and advocating for greater recognition of regional data, languages, and journalistic standards in global AI systems.

Toward Institutional Governance, Not Individual Coping
Across the discussion, a recurring theme emerged: journalists are currently left to individually adapt to AI, often without clear guidelines, training, or institutional safeguards. This ad hoc adaptation is unsustainable.

Panelists stressed the need to move from individual survival strategies toward institutional governance frameworks, including newsroom policies on AI use, transparency standards, data sourcing guidelines, and ethical oversight.

Without such structures, AI risks reshaping journalism into a form of prompt engineering detached from contextual understanding and public accountability.

The Road Ahead
Arab journalism stands at a crossroads where AI tools that can enhance efficiency, expand access, and support investigative work, yet without intentional governance, linguistic investment, and regional advocacy, these same tools may deepen dependency, distort knowledge production, and erode trust.

The way forward lies not in rejecting AI, but in shaping it, through collective action, stronger institutions, and a renewed commitment to digital sovereignty. As the voices in this dialogue made clear, responsible AI in Arab newsrooms is not a technical challenge alone; it is a cultural, political, and ethical project that demands regional leadership and global engagement.