Dubai RTA rolls out AI-only driving test result system

What happened

Dubai's Roads and Transport Authority (RTA) has rolled out a new generation of driving tests that removes human examiners from the assessment of results, relying instead on artificial intelligence to determine pass or fail outcomes. The system combines facial recognition, computer vision and 3D mapping to track a candidate's behaviour and vehicle positioning throughout the test, with automatic emergency braking built in as a safety layer.

According to Arabian Business, the technology is designed to standardise how tests are scored by removing subjective human judgement from the final outcome, while still monitoring the candidate's actions in real time during the drive.

Why it matters

This is fundamentally a story about what AI-enabled sensing and computer vision now make possible in a public service that has traditionally relied on individual examiner discretion. By automating both observation and adjudication, RTA is signalling a shift toward outcome consistency and auditability in a process that citizens and residents interact with directly and often repeatedly.

For transformation leaders, the move illustrates a broader pattern in GovTech: pairing perception technologies (vision, mapping, biometrics) with rules-based decisioning to reduce variability in high-volume, high-stakes public services. It also raises the operational bar — any AI system making pass/fail determinations needs to be transparent, defensible and trusted by the people it assesses, not just technically accurate.

The Renascence take

Removing the examiner from the results loop solves a real fairness problem — inconsistent scoring — but it introduces a new one: perceived fairness. Candidates who fail an AI-scored test may find it harder to accept a verdict they cannot argue with in the way they might have contested a human examiner's call.

The behavioural risk here isn't accuracy, it's legitimacy. People forgive human error more readily than machine error, because a person can be reasoned with and a system feels final. If RTA wants candidates to trust an AI verdict as much as a human one, the design challenge is less about the sensors and more about explainability — giving people a clear, specific reason for the outcome, and a visible path to review it. Operators rolling out algorithmic adjudication anywhere in the citizen journey should treat the explanation layer as seriously as the detection layer.