Dubai, UAE – The UAE’s Artificial Intelligence, Digital Economy, and Remote Work Applications Office announced that Google Gemini has ranked first in the “AI in the Ring” Index, the world’s first benchmark designed to evaluate how effectively AI language models reflect Emirati culture, dialects, traditions, and national values through a challenge centered on cultural intelligence within the UAE context.
Gemini earned the top ranking following a review of over 400 questions and 5,200 generated responses from 11 major language models. A Committee of Emirati experts evaluated the outputs to identify which models demonstrated the strongest cultural understanding.
Following Gemini 2.5 Pro, the list of the top five high-performing models included: ChatGPT (OpenAI GPT-4o), ChatGPT (OpenAI o1), Cohere, and Grok.
Omar Sultan Al Olama, Minister of State for Artificial Intelligence, Digital Economy, and Remote Work Applications, emphasized that national identity remains a priority for the UAE, especially as digital transformation accelerates. He noted that emerging technologies must reflect and reinforce the country’s authentic values, ensuring that AI systems enhance save the UAE’s cultural heritage forward with clarity and respect.
Al Olama stated that the index reinforces a core national principle: technology should grow from a nation’s roots. It should protect identity, deepen cultural awareness, and shape a future where innovation advances confidently in harmony with the spirit, character, and enduring traditions of the United Arab Emirates.
The assessment measured each model’s ability to reflect the culture of the UAE while preserving language, heritage, and values. It involved a direct comparison between international models to evaluate their understanding of Emirati identity across seven dimensions, including Historical context, Creative and poetic writing, Emirati Arabic, language and dialects, Cultural symbols and meanings, Social context and behavioral norms, Social and religious topics and Emirati values and ethics.
The models were tested using a Human–AI red-teaming framework, intentionally challenging them to highlight cultural gaps, risks, or misinterpretations. A custom-built platform monitored bias and unusual outputs, and all model identities were concealed from the review committee to ensure neutrality.
Underscoring the urgency of culturally grounded AI, less than 5% of Arabic content worldwide exists in digital form, while 48% of Generation Z now relies on AI as their primary source of information, a reality that makes training models on accurate, UAE-relevant cultural data essential to safeguarding national heritage for future generations.
