Where Self-Driving Cars Really Stand Today

Sjoerd van der Wal/GettyImages

Predictions that fully self-driving cars would be everywhere within a few years have been made, and missed, many times. Yet driverless taxis are now carrying paying passengers in several cities, while many new cars can steer and brake themselves on highways under supervision. The picture is confusing partly because "self-driving" is used to describe very different technologies. The industry's levels of automation are the clearest way to understand what exists today.

The levels of driving automation

SAE International, an engineering standards body, defines six levels, from 0 to 5:

  • Level 0: no automation. The car may give warnings, but the driver does everything.
  • Level 1: the car helps with steering or speed, such as adaptive cruise control, but not both at once.
  • Level 2: the car can steer and control speed together in some conditions, but the driver must supervise constantly and be ready to take over at any moment.
  • Level 3: the car drives itself in specific conditions, and the driver can take their attention off the road, but must take over when the system asks.
  • Level 4: the car drives itself with no human needed, but only within a defined area or set of conditions.
  • Level 5: the car can drive itself anywhere a human could, in any conditions. No such vehicle exists.

Level 2: the systems most drivers can buy

Most driver assistance features on sale today are Level 2. Examples include Tesla's Autopilot and Full Self-Driving (Supervised), GM's Super Cruise and Ford's BlueCruise, which allow hands-free driving on mapped highways while monitoring the driver's attention with a camera. Despite their names, all of these require an attentive driver. Investigations by the US National Highway Traffic Safety Administration into crashes involving driver assistance systems have highlighted the danger of drivers relying on them too heavily.

Level 3: limited and conditional

Mercedes-Benz's Drive Pilot became the first Level 3 system approved for use on certain public highways in the United States, receiving approval in Nevada and California in 2023. It works only in specific conditions, such as dense traffic on approved highways at limited speeds, during daylight and good weather. When conditions change, the driver must take back control.

Level 4: robotaxis in defined areas

The most advanced autonomy today is found in robotaxi services that operate without anyone in the driver's seat, within mapped service areas. Waymo, owned by Google's parent company Alphabet, runs fully driverless paid rides in several US cities, including Phoenix, San Francisco and Los Angeles, and has been expanding to more. In China, companies such as Baidu, with its Apollo Go service, operate driverless robotaxis in cities including Wuhan. Tesla began a small robotaxi service in Austin, Texas, in 2025, initially with safety monitors on board.

The industry has also seen setbacks. In October 2023, a Cruise robotaxi dragged a pedestrian who had been thrown into its path by another vehicle, and California regulators suspended the company's driverless permits. General Motors later ended its funding of Cruise's robotaxi business in December 2024.

Are they safer than human drivers?

Supporters argue that automated systems don't get tired, distracted or drunk, which are factors in many crashes. Waymo has published studies, based on millions of driverless miles, suggesting its vehicles are involved in fewer injury-causing crashes than human drivers in the same cities. Independent researchers note that comparisons are complex, because robotaxis operate in limited areas and conditions. Overall, the evidence for Level 4 services in their operating areas is encouraging, but it does not yet show how they would perform everywhere.

What is holding back wider adoption

  • Edge cases: unusual situations, such as emergency scenes, construction zones, police directing traffic and unpredictable pedestrians, remain challenging.
  • Weather: heavy rain, snow and fog can interfere with cameras and sensors.
  • Cost: the sensors and computing used in robotaxis are expensive, though costs are falling.
  • Regulation: rules differ by state and country, and questions about liability in crashes are still being settled.
  • Public trust: surveys by groups such as AAA have found that many drivers remain wary of fully self-driving vehicles.

What this means for drivers today

If your car has driver assistance features, read the manual to understand exactly what they can and cannot do, and keep your eyes on the road and hands ready. If you live in a city with robotaxis, riding in one is the easiest way to see the technology at work. Fully self-driving cars that you can buy and send anywhere remain a longer-term prospect, but driverless services within defined areas are already part of everyday life in a growing number of places.