The Self-Driving Car: What It Promises and What It Does Not Guarantee

The number that drives the idea

The World Health Organization says about 1.16 million people die each year in road crashes, that these crashes are the leading cause of death for those aged 5 to 29, and that 92% of deaths occur in low- and middle-income countries. Figures like these make any technology that promises to reduce them worth considering.

It is often said that 94% of crashes are caused by humans. The source is a study by the US National Highway Traffic Safety Administration (NHTSA) that tied the "critical reason" to the driver in about 94% of the crashes studied. But the report itself cautions that the "critical reason" is the immediate reason for the pre-crash event, not the cause of the crash and not an assignment of blame to the driver. The figure points to room for improvement; it does not acquit the road, the vehicle and the regulation.

The US agency classifies automation from Level 0 to Level 5. At Levels 1 and 2 the driver remains responsible for monitoring. At Level 4 the system drives with no engaged driver, but only within limited service areas. At Level 5 it drives in all conditions and on all roads. So the phrase "self-driving car" can mean very different things, and the difference between a driver-assistance feature and a driverless taxi is a difference in kind.

What do the data say?

Waymo publishes safety figures for its driverless vehicles. Up to June 2026 they had covered about 271 million miles, and compared with a human benchmark it estimated about 82% fewer injury-causing crashes and about 95% fewer serious-injury-or-worse crashes. The comparison rests on police data in Phoenix, San Francisco, Los Angeles and Austin, excluding freeways. The company says its methods have been published in peer-reviewed papers.

These are serious results, but they have limits: the source is the company itself, the scope is particular cities and driving conditions, and transfer to Gulf streets and traffic patterns needs local testing, not assumption.

What the numbers do not settle

• Responsibility: when the vehicle errs, who answers: the maker, the operator or the owner? This is a matter for legislation, not engineering.
• Central failure: when a fleet shares one piece of software, it may share its vulnerability, and a networked vehicle is a potential target for intrusion. Cybersecurity must therefore be part of the design from day one.
• Privacy: a city that knows the location of every vehicle at every moment knows where people go and when. Governing the collection of this data is settled by regulation, not factory settings.
• Livelihoods: driving supports many families, and slogans about "retraining" do not suffice to address the transition's effect on them; realistic transition paths that precede deployment are needed.
• Fairness: the service may reach the easier neighbourhoods first, leaving those who need it most at the back of the queue.

Riyadh: a trial under supervision

According to Arab News, a self-driving taxi trial began in July 2025 between King Khalid International Airport and central Riyadh, in a partnership between WeRide, Uber and a local partner. The report said WeRide was the first autonomous-vehicle company to complete the Transport General Authority's regulatory sandbox, and that a self-driving vehicles code requires smart road-communication devices. An approach built on limited trials and supervision before expansion is what the technology needs at this stage. On the dream of intersections without signals in the twin city, see the article "Roads Without Signals: What the Twin City Promises and What It Does Not".

What is to be done

The practical conclusion comes down to four principles: cautious, gradual progress through limited trials before expansion; independent auditing of safety results rather than reliance on what companies announce; legislation that precedes deployment and sets responsibility and data limits; and thinking about transport as a whole, since the larger gain may come from shared fleets that reduce congestion, not from a private car that drives itself. The aim is that everyone arrives safely, that we give thanks for the blessing by protecting lives, and that we take the means to protect them.