AI incidents are reaching senior leadership in UAE organisations as companies move from experimentation toward more autonomous systems, according to a study commissioned by IBM and conducted by Censuswide. The research surveyed 3,221 executives globally, including respondents in Saudi Arabia and the UAE. In the UAE results, 75 percent of executives said an AI-related incident had reached the board or C-suite during the previous year.
The findings point to governance becoming an operational concern rather than a policy exercise. Forty-nine percent of UAE respondents said their organisation had changed a governance policy or framework in response to AI incidents. At the same time, the survey found that only 39 percent had formal board-reviewed processes for quantifying financial risk associated with AI. These figures describe respondent perceptions and practices; they are not an audit of every organisation in the UAE.
The study also records expectations about agentic AI. Seventy-eight percent of UAE executives expected some degree of agent autonomy by 2027. An autonomous agent can retrieve information, use tools and perform bounded actions across business systems, which means the risk is shaped by identity, permissions, data access and escalation design as well as by model quality. A model can produce a plausible answer while an improperly configured workflow still creates an unacceptable action.
IBM reported that 60 percent of UAE respondents were continuing with AI projects within agreed risk limits while governance adapted, while 30 percent were delaying projects until governance was ready. That split suggests organisations are making different judgements about acceptable exposure, control maturity and the strategic cost of waiting. It also makes the evidence trail important: leaders need to know which systems are deployed, what they can do, how they are monitored and who can stop them.
The commercial implication is a growing need for model evaluation, AI asset inventories, access controls, incident response, policy enforcement and independent assurance. The study does not prove that UAE companies are uniquely exposed, but it does show that the country’s adoption conversation is reaching the boardroom. As agent autonomy grows, governance that is measurable, tested and connected to real operating permissions will matter as much as model capability.
