The United Arab Emirates has established a clear boundary regarding the application of artificial intelligence (AI) in the public sector: no AI agents will be authorized to make decisions that could affect a person's livelihood.
Omar Sultan Al Olama, Minister of State Council for Artificial Intelligence at the Prime Minister's Office, emphasized that this rule is particularly relevant now because the true capabilities of the technology have not yet been fully studied. He stated that human life or human well-being are priceless and therefore cannot be assigned a monetary value.
These statements were made by him during a panel discussion titled 'Trust Test for Agentic AI' at the annual World Economic Forum Global Leaders meeting in Dubai on Wednesday. At the event, he also answered questions about global regulation, public concerns regarding data centers, and the threat of job displacement.
The rapid growth of AI has caused international concern regarding its impact on employment, as millions of workers fear that automation may replace them in the workplace. While AI promises increased productivity and efficiency, experts warn of a growing risk to routine professions, highlighting the urgent need for reskilling, education, and workforce adaptation.
Al Olama reported that UAE leaders aim for AI agents to provide 50 percent of federal government services within two years. After mapping all government services, it was determined that about 70 percent could allow for certain uses of AI. However, in the remaining 30 percent, the risk of error exceeds the acceptable level for the government, so these services are not yet considered for automation.
He noted that low-risk tasks, such as passport renewal, are suitable for agent work. Services requiring a higher level of verification of agent actions are currently excluded.
Al Olama also introduced a concept he termed the first public mention of an 'agent control tower' on the federal government side. Under this system, every agent must connect to it, and every action and 'thought' of the agent is logged.
AI Agents undergo auditing
Government agents are subject to continuous auditing by both other agents and humans. Furthermore, government agents are only implemented by UAE entities that bear responsibility and understand the consequences, rather than those who purchase them from companies lacking such understanding.
Regarding global rules, Al Olama believes that debates should not be limited to slowing down or accelerating AI development, as, in his words, 'this train has already left the station,' referencing the declared determination of the US President not to fall behind China. Instead, he called on countries to first agree on the most extreme risks, starting with autonomous weapons capable of making lethal decisions. He noted that few dispute that such decisions should not be left to machines, but there is no global consensus on this matter.
He compared this situation to decades of attempts to limit nuclear weapons, stating that if one side agrees and the other does not, it puts it at a disadvantage. Al Olama urged governments to set a firm deadline for global AI regulation. With nine years of experience as an AI minister, he does not want to look back in nine years and see no progress.
He mentioned that warnings about ChatGPT-2 being too dangerous for public release emerged in the early 2020s. Since then, the UAE has participated in discussions at G20, G7, and UN forums but has not reached a consensus. Al Olama explained that the fear stems from a lack of understanding of the technology. The UAE has focused on education, including the annual AI Day on October 29th, when every resident receives an SMS with an introductory prompt.
He also clarified that data centers in the UAE do not increase energy prices because the country utilizes solar and nuclear energy, as well as having reserve capacity. Additionally, he reported that the UAE does not have natural water sources for grid intake and relies on desalination.
Concerning employment issues, he stated that the program to train one million people as 'super-prompters' will conclude this month. Every federal employee must undergo extensive AI training by August of next year, with half of the employees having already completed it. He concluded that the UAE model is unique and cannot be directly copied in the US, China, or India, although some elements can be adapted.
}))