DUBAI, United Arab Emirates — Emirates NBD and OpenAI have announced an enterprise collaboration to bring advanced artificial intelligence models into the bank's operations using computing infrastructure located in the United Arab Emirates, as financial institutions seek ways to adopt AI without compromising sensitive data or regulatory requirements.
The agreement is expected to make Emirates NBD the first bank in the Middle East and North Africa to deploy OpenAI's GPT models using in-country inference infrastructure. The arrangement will allow the bank to introduce AI capabilities across customer services, risk management, compliance and internal operations.
Unlike conventional arrangements that may involve processing information on infrastructure outside a bank's home market, the collaboration is designed to support AI processing within the UAE. The approach addresses an important consideration for financial institutions, which must balance technological development with obligations surrounding customer information, operational security and regulatory oversight.
A central feature of the collaboration is the use of in-country inference infrastructure. Inference refers to the process through which a trained AI model analyzes information and generates responses or predictions when a user or software system submits a request.
Locating that processing infrastructure within the UAE is intended to help Emirates NBD manage where information is handled while adopting OpenAI's models across its business. Data residency is particularly relevant to banks because their systems routinely process financial records, customer information and other sensitive material.
The companies describe the arrangement as part of a broader move toward sovereign AI, an approach that emphasizes local control over certain aspects of AI infrastructure and data processing. In this case, the emphasis is on where models perform inference rather than on developing an entirely new AI model within the UAE.
Local inference infrastructure does not, by itself, establish that every element of an AI service remains within national borders. Questions about data storage, system access, model management and operational safeguards still depend on how individual services are configured.





