Emirates NBD, the Dubai-based banking group, announced on October 19, 2026 that it will deploy OpenAI's GPT models using in-country inference infrastructure in the UAE. The bank and OpenAI call it a first for a bank in the Middle East and North Africa (MENA).
The deal is light on numbers. Neither side gave a price, a model name or a rollout date. But it is a clear example of how regulated industries now buy frontier AI: the question is no longer only "which model," but "where does the model run, and what leaves the country."
What Emirates NBD says it will do
The press release lists four areas where the bank will deploy OpenAI's models:
- Customer engagement across digital channels
- Risk and compliance, with advanced analytics
- Automation of knowledge-intensive workflows
- Employee productivity
The bank also says it will connect OpenAI's models to its internal systems, APIs and enterprise data. It expects that to speed up new digital services, improve customer personalisation, raise engineering and operational productivity, and strengthen risk monitoring and financial crime controls.
Miguel Rio-Tinto, Group Chief Digital and Information Officer at Emirates NBD, said in the release: "AI is not an experiment for us, it is a core capability that will define how we build, serve and scale." Farouk El Hamzawi, Head of Enterprise at OpenAI MENA, said OpenAI will support the bank "with OpenAI models, tools and UAE-based inference, helping the bank meet regulatory requirements."
For scale, Middle East AI News reports that Emirates NBD operates in 13 countries, serves over 25 million active customers and had total assets of AED 1.3 trillion at June 30, 2026. The same outlet says no financial terms were disclosed.
What "in-country inference" means
Inference is the step where a model reads a request and writes an answer. In OpenAI's terms, data residency is about storage: where content such as conversations and files sits at rest. Inference residency goes a step further. The GPUs that process the prompt are also in the region.
The release frames this as the point of the deal. It says in-country inference "allows financial institutions to adopt frontier AI at enterprise scale while supporting local regulatory expectations around data residency, privacy and operational resilience."
OpenAI's public terms for this are on its help page for ChatGPT data residency and inference residency. As of October 19, 2026, that page says:
- Inference residency for ChatGPT is available in three regions: Europe (EEA plus Switzerland), the United States and the United Arab Emirates.
- It is offered to eligible ChatGPT Enterprise and Edu customers, and it requires data residency to be turned on in the same region first.
- With it enabled, GPU inference on in-scope customer content (prompts, files, conversations and embeddings derived from them) runs on GPUs in the selected region.
What still may leave the region
The same help page is direct about the limits. Inference residency "does not extend to all CPU-based processing or to system data more broadly." Across two sections, it lists what may still be processed or stored outside the selected region:
- CPU-based processing, such as extracting text from uploaded PDF or Word files, routing requests and coordinating steps across services
- Authentication, routing and analytics
- Anything sent to an external app, MCP server, web browsing provider or other third-party service, which falls under that provider's own terms
- Some stored data, such as billing information, user logins and workspace metadata
Account data, billing details, high-level usage statistics and logs without content are out of scope. The page also answers its own question, "Does inference residency guarantee that all processing stays in-region?", with "No."
For the UAE specifically, the page lists feature limits. GPT-5.5 and GPT-5.6 Luna are the models available in ChatGPT with UAE inference residency. Image generation, internal search, ChatGPT Work and improved memory are not supported, and GPT-Live is not available for those workspaces.
One caution: these are the published terms for ChatGPT. The Emirates NBD release does not say whether the bank will use ChatGPT Enterprise, the API, or both, and OpenAI's help page points API customers to separate API data residency documentation. So the ChatGPT limits above show how OpenAI defines the service, not a confirmed description of the bank's setup.
Why a bank cares
The release says banks operate in a highly regulated industry "where security, governance and compliance are non-negotiable." In-region GPU execution answers one question regulators and risk teams ask: where the model processes customer content. It does not answer all of them. A bank still has to decide which workflows send customer content to the model, which tools and connectors are allowed, and how outputs are checked before they reach a customer or a compliance file. The help page's list of out-of-region processing is the list a bank's risk team will read most closely.
The four use areas in the release also differ in risk. Employee productivity tools stay inside the bank. Customer engagement and financial crime controls touch customers and regulators directly. The release gives no timeline for any of them. PCMag Middle East noted the same gap: the release names no date for customer-facing features.
How this fits OpenAI's UAE build-out
OpenAI has been building in the UAE for over a year. In May 2025 it announced Stargate UAE, its first partnership under the OpenAI for Countries program. Middle East AI News reports that OpenAI added the UAE as a ChatGPT data residency region in November 2025 and switched on UAE inference residency in August 2026.
The Emirates NBD release does not name the data center or infrastructure behind its in-country inference, so it is not confirmed that the bank's workloads run on Stargate UAE. What the release does confirm is the offer OpenAI is making to regulated buyers in the UAE: its models, with inference run in the country.
Middle East AI News calls Emirates NBD the second largest banking group in the UAE and "a great first reference customer" for OpenAI's inference residency in the region.
What to watch
- Which models and products the bank actually uses (ChatGPT Enterprise, the API, or both)
- The first customer-facing feature and its launch date
- Whether the risk and financial crime work moves past analytics into automated decisions
- Whether other MENA banks sign similar in-country deals, and whether OpenAI adds more inference residency regions, which its help page says it plans to do
Sources
- Emirates NBD: Emirates NBD Becomes First MENA Bank to Harness OpenAI's Sovereign AI Infrastructure for customers in the UAE
- OpenAI Help Center: Data residency and inference residency for ChatGPT
- Middle East AI News: Emirates NBD first MENA bank on OpenAI's in-country inference
- PCMag Middle East: Emirates NBD to Run OpenAI's GPT Models on UAE-Based Infrastructure
- OpenAI: Introducing Stargate UAE
- Gulf News: Emirates NBD becomes first MENA bank to run OpenAI models in UAE
