UAE weighs AI agent risks as automation expands in government

AI security experts warn that AI agents operating on behalf of employees with legitimate credentials create new accountability risks, as determining who authorized specific actions becomes increasingly difficult when automated systems make unexpected decisions. Traditional cybersecurity focuses on whether users or processes are malicious, but agentic AI requires organizations to also assess whether an AI should be permitted to take particular actions. The US National Institute of Standards and Technology recommends treating agents as "first-class entities" with dedicated identities, while Microsoft advises limiting access and recording full activity chains to answer what happened, under whose authority, and what changed. The issue carries particular weight in the UAE, where the federal government targets transitioning 50 percent of government sectors to agentic AI within two years. Experts argue that control requires visibility into AI actions and the ability to intervene before high-risk decisions, balancing automation speed with human judgment.