Shifting to the Post-AI Models Era

Talal Abu-Ghazaleh

Concerns are increasing over the dominance of Artificial Intelligence (AI) systems developed in the United States, which primarily serve Western societies and English-speaking audiences.

On the other hand, other countries are seeking to accelerate the development of their AI systems through domestic models based on national data and technologies, aiming to achieve what is known as "AI Sovereignty". This means reaching full self-reliance across all stages of the AI lifecycle.
In response to this trend, India launched the "India Stack", emphasizing the necessity of developing national AI solutions that meet local requirements. India, therefore, seeks to achieve technological sovereignty by fostering domestic innovation throughout the AI lifecycle while reducing reliance on foreign technologies. This is to strengthen its control over its digital future through building and hosting local AI models to meet national needs.

Shifting toward AI sovereignty is increasingly influenced by the policies of the current US Administration, which is redrawing traditional alliances with a focus on national interests at the expense of international cooperation. Under this approach, non-US societies; especially those with limited user bases or those outside the English-speaking world, may find themselves marginalized regarding content moderation and AI technology development. Such concerns have encouraged countries to develop domestic solutions and re-evaluate their reliance on the United States, which is effectively excluding many communities around the world.

The current dominance of Anglo-Saxon datasets in training AI models contributes to clear disparities in performance, especially for non-Western languages. Multilingual models show poor performance when processing languages such as Chinese and Hindi, highlighting the critical importance of providing diverse and comprehensive datasets that reflect linguistic and cultural multiplicity. In this context, localized language models offering a promising approach to overcome these challenges and bridge existing gaps.

I therefore urge all capable countries in the Middle East to build and develop their AI models, relying on datasets that reflect the distinctiveness of this region. This will effectively contribute to preserving its language, history, and culture, ensuring they are documented in the record of contemporary technology to become an essential part of this modern tech revolution. We have looked into the past for too long, admiring the great achievements made by Arabs throughout their history. Today, it is time to shift toward building societies that actively contribute to improving the current global status quo.

I believe that it is necessary to develop alternatives to US AI models, so we can shape a future where technology serves as a tool for global advancement, rather than just for the advancement and benefit of a single group of people.