The UAE Is Treating AI Like Infrastructure. Other Governments Should Too
Zeta42 · 20 September 2026
At Dreamforce, UAE AI minister Omar Sultan Al Olama said data centre locations will no longer be promoted. His wider approach, from older models to mass training, is the right one.
The data centre is the new bridge
Omar Sultan Al Olama, the UAE's Minister of State for Artificial Intelligence, Digital Economy and Remote Work Applications, made a point at Dreamforce in San Francisco that deserves wider attention. The UAE will change its requirements for data centres and stop promoting where they are. Data centres, he told Salesforce chief executive Marc Benioff, have become critical infrastructure and a target.
"Unfortunately they're more important than bridges. They're more important than roads. They're more important than tunnels because our whole lives are on these devices that we're using."
He is right, and the point reaches further than security policy. In a country where most government services now use large language models to answer questions, AI is no longer a showcase technology. It is plumbing, and it deserves the same protection and upkeep as the roads it now outranks.
Why the locations are going quiet
The conversation with Benioff turned to how exposed physical infrastructure has become. Al Olama's answer was blunt: "We're not going to promote the locations of data centers for anyone who wants to have any idea to do something with them." He expects most countries will need to change their rules too.
Anyone who still thinks of AI as an app on a phone should sit with that. The models run somewhere physical, and where they run has become a question of national security.
Older models, used well
The most useful part of Al Olama's argument was about models, not buildings. Governments, he said, do not need the newest models to improve people's lives. Most UAE government services now use large language models to answer questions, and they use older models rather than frontier ones. His example was TAMM, an Abu Dhabi app where users tell an agent what they need and it handles the whole government process for them.
He also drew red lines. If the consensus is that frontier models carry cyber security risks, he said, governments should agree not to deploy them, because a leak or the loss of records would affect millions of people. He rejects AI that predicts crime because of false positives: a person the system flags might never commit the crime. And he warned against a surveillance state, saying the UAE looks at AI for quality of life, not economic gain.
This is the part other governments, and plenty of companies, should copy. The right model is the one that does the job safely, and some uses should stay off the table entirely. That is a more grown-up stance than a race to the frontier.
Train people, and don't let fear decide
Al Olama, who became the UAE's AI minister in 2017, put all of this in a long historical frame. He said the Middle East led the world in technology, economy and culture from 813 AD to 1515 AD, then banned the Gutenberg printing press for 200 years and fell behind. He traced the ban to fears of misinformation and of the unknown, and compared those fears to today's calls to regulate AI. "So, we can't afford to make that mistake again. And even if you fear certain things, you have to put your head down and focus on the things that you know are going to change people's lives," he said.
The UAE's answer to that fear has been training. Al Olama said the country has trained 80,000 people in agentic AI, every employee of the federal government, after first training a million people to write prompts for AI. He expects AI to create jobs rather than destroy them. On regulation, he described how the UAE wrote its data protection law: instead of copying the EU's GDPR, it asked data companies to draft the law, then government and civil society filled the gaps.
What it asks of the rest of us
We would add one caution, and Al Olama raised it himself. He warned of an age of "incredible inequality" and said companies must make the technology fair and affordable for Africa, Asia and Latin America. The same logic applies inside a single country. The benefits go to the people who know how to use the tools.
That is why the 80,000 figure is the one we would underline. Keeping data centre locations quiet protects the infrastructure. Training people is what makes it worth protecting. At Zeta42 we treat AI literacy as infrastructure too: knowing which tool fits a task, and where its limits are, matters more than knowing the name of the newest model.
Source: thenextweb.com
