UAE Grapples with AI Accountability as Autonomous Systems Enter Federal Decision-Making

GOVERNANCE CHALLENGE EMERGES AS UAE PUSHES AUTONOMOUS AI INTO FEDERAL OPERATIONS

The United Arab Emirates government faces a threshold question that no state has yet resolved: what decisions can be delegated to an autonomous machine, and who bears responsibility when those decisions fail. That question moved from theoretical to operational this month when more than 100 federal officials gathered for a workshop launching the strategic phase of the UAE’s agentic AI project, a two-year initiative to convert half of all federal government operations, services and tasks to autonomous AI models.

The National Committee for the Agentic AI Project convened the session to establish implementation priorities, timelines and the frameworks that will determine which government tasks qualify for autonomous execution at all. That classification exercise is the crux of the entire programme. The UAE possesses the computational infrastructure and data architecture most governments are still building: an AI-powered performance system tracking more than 150 million data points monthly, a sovereign cloud, digital identity systems and data-sharing layers. What remains absent is a published, defensible standard for distinguishing tasks an autonomous system may complete from those it may only recommend.

The UAE’s institutional head start in artificial intelligence spans nine years. The government published a national AI strategy in October 2017 and established a ministerial post to oversee it days later, making Omar Sultan Al Olama, then 27, the world’s first minister of state for artificial intelligence. That institutional foundation now faces its most complex test: whether the infrastructure and strategy translate into workable governance when machines begin making autonomous decisions across federal agencies.

The workshop operated under a stated principle: human leads, AI enables. Yet the founding language used to introduce the project describes something different. When announcing the framework, Sheikh Mohammed bin Rashid Al Maktoum said the UAE would be “the first government in the world to largely deploy Agentic AI models” across sectors and operations, describing systems that manage operations and execute independent series of actions without human intervention. The gap between autonomous execution and human oversight is precisely what the classification frameworks are designed to close, but until those frameworks are published, accountability remains undefined.

Huda Al Hashimi, Deputy Minister of Cabinet Affairs for Strategic Affairs, characterized the workshop as the operational beginning of deploying AI models across government activity. The programme spans seven pillars: strategy and projects, foresight and strategic intelligence, policies, structures and governance, government performance, global competitiveness and innovation in government work. Officials also discussed performance indicators and mechanisms to prevent entities from duplicating each other’s deployments.

Oversight responsibility rests with Sheikh Mansour bin Zayed Al Nahyan, with a taskforce chaired by Mohammad Al Gergawi, Minister of Cabinet Affairs. In May, the Cabinet approved the first package of government services to receive agentic AI tools and authorized the largest training programme in UAE government history, covering 80,000 federal employees from ministers to new staff. The same session approved a national AI healthcare policy and frameworks designating digital records as the official source of core government data, with requirements for single collection and secure sharing across entities.

The implementation timeline has been deliberate. A Dubai training event in June drew more than 300 participants from 50 federal institutions, with a 90-day goal of identifying which services and procedures suit agentic AI. A separate workshop that month, conducted by the Presidential Court and Ministry of Cabinet Affairs, trained 600 staff through identical material.

By contrast, the material released so far addresses pillars, training, indicators and governance structures but does not describe recourse when an autonomous system makes an error affecting a citizen, nor does it clarify where liability lands when decisions go wrong. That silence is not unique to the UAE. No government has answered these questions because no government has deployed agentic AI at scale before.

The difference is that the UAE will have to answer first, on a two-year timeline, with public evidence about what agentic AI in government actually does to service delivery, headcount and error rates. Governments across Asia and Europe are approaching the same decision from the beginning, most with an AI strategy and a national body but no quantified targets or deadlines. The UAE’s numerical commitment and fixed date create an experiment others can observe without conducting themselves, though they must eventually decide whether to act before results exist.

The task classification frameworks represent the document to watch. When published, they will constitute the first serious attempt by any state to write down which decisions a government is willing to let a machine make on its behalf. Everything else in the programme is procurement and training. That classification exercise is constitutional. Whether the frameworks, once published, prove adequate to the accountability questions they are meant to settle is the test no workshop can answer in advance.