The 6 Official Levels of Autonomous Vehicles — And Where Today’s Cars Fit In

For decades, self-driving vehicles have been a dream of science fiction, a vision of a world where cars drive us around while we relax in the back seat. In reality, turning that dream into a safe, reliable product is proving far harder than it looks. Automakers and tech companies have invested billions in developing driver-assist systems and artificial intelligence platforms, yet fully autonomous cars remain limited to small-scale pilot programs.

One reason for the confusion is that “self-driving” can mean very different things depending on who you ask. To bring clarity, the Society of Automotive Engineers (SAE) and the U.S. Department of Transportation defined six clear levels of automation, ranging from Level 0 (no automation) to Level 5 (complete autonomy under all conditions). These standards are now used worldwide to describe the level of control a vehicle can have.

Let’s break down each level, what the capabilities really mean, and which cars currently fit into each category.

Level 0: No Driving Automation

Level 0 means the driver is entirely in control at all times. The vehicle may provide alerts or short interventions, such as lane departure warnings, blind spot monitors, or automatic emergency braking, but these do not constitute automation because they don’t consistently control the vehicle.

This is still the most common category on U.S. roads. Older vehicles without adaptive cruise control or steering assistance remain firmly at Level 0. Even modern cars with only passive safety alerts, like a 2010 Honda Accord with stability control, are still Level 0 under SAE’s definitions.

Level 1: Driver Assistance

Level 1 adds limited automation in one area: either steering or acceleration/braking, but not both simultaneously. The driver is still responsible for monitoring the road and handling all other tasks.

A familiar example is adaptive cruise control, which automatically adjusts the car’s speed to maintain a safe distance from the traffic ahead. Another is lane-keeping assist, which gently corrects steering if you drift.

Typical vehicles Include Many mainstream models, such as the Toyota Camry with Toyota Safety Sense, the Honda Civic with Honda Sensing, and the Ford Escape with Co-Pilot360. These systems reduce fatigue on long trips but still require both hands on the wheel.

Level 2: Partial Driving Automation

Level 2 represents the first big step toward autonomy. These systems can simultaneously control steering, acceleration, and braking under certain conditions, such as highway driving. However, the driver must remain engaged, monitoring the road at all times, and be ready to take over instantly.

In practice, Level 2 cars can maintain their position in a lane, adjust speed automatically, and even execute lane changes; however, the human driver remains legally and practically responsible for the outcome.

A few modern cars have Level 2 capabilites:

• Tesla Autopilot and Full Self-Driving (FSD) Beta (still classified as Level 2, according to regulators, as drivers must still supervise) are the best examples of the capabilities of these vehicles.
• GM Super Cruise (available on Cadillac Lyriq, Chevrolet Silverado EV, and GMC Hummer EV).
• Ford BlueCruise (on Mustang Mach-E and F-150 Lightning).
• BMW Driving Assistant Professional and Mercedes Driver Assistance packages.

Most new premium and even mid-market vehicles in the U.S. are now shipping with some form of Level 2 driver assistance.

Level 3: Conditional Driving Automation

Level 3 is where the definition of “self-driving” starts to feel real. At this stage, the car can manage all driving tasks, steering, speed, braking, and lane changes under limited conditions, such as clear weather on mapped highways. Importantly, the driver can take their eyes off the road and hands off the wheel.

But there’s a catch: the system may at any time ask the driver to retake control, and the driver must be able to do so quickly. This makes Level 3 much harder to implement legally, since regulators worry about “handover” safety if drivers are distracted.

Typical vehicles:

• Mercedes-Benz Drive Pilot is approved for use in Germany and in limited parts of California and Nevada.
• The Honda Legend, offered in Japan with a leasing program, is the first production car to legally feature Level 3 automation.
• Zeekr 9X SUV, announced in March 2025, which will ship with Level 3-ready technology, including advanced LiDAR and Nvidia Drive Orin processors. Deliveries are expected later this year in China pending regulatory approval.

Level 3 remains rare, but it marks a crucial milestone for cars capable of “self-driving” in certain real-world scenarios.

Level 4: High Driving Automation

At Level 4, vehicles can drive themselves entirely within specific geofenced areas or conditions, such as well-mapped city zones, without needing human intervention. Unlike Level 3, if something goes wrong, the vehicle is expected to handle the situation safely without requiring driver takeover.

This is the level at which robotaxis operate today. Services operated by Waymo and Cruise in cities such as Phoenix, San Francisco, Los Angeles, and Atlanta are considered Level 4. Passengers can hail a ride with no driver in the front seat.

However, Level 4 cars cannot yet drive anywhere, anytime. They are limited to mapped zones, fair weather, and specific regulations.

Level 5: Full Driving Automation

Level 5 is the ultimate goal: vehicles that can drive anywhere, in any conditions, with no steering wheel, pedals, or human input required. At this level, every occupant is a passenger.

So far, Level 5 remains theoretical. No production vehicle in the world today offers it, although companies like Tesla, Waymo, Zoox, aim to achieve it eventually.

Why Full Autonomy Is So Difficult

Moving from Level 2 to Level 5 isn’t a matter of flipping a switch. Engineers face enormous challenges, such as snow, construction zones, and unpredictable human drivers, which create scenarios that even the smartest AI struggles with.

One of the biggest roadblocks to the development of fully autonomous vehicles remains the cost. Key components like LiDAR sensors, radar units, and advanced computing systems can cost tens of thousands of dollars on their own, even before integration. Once you factor in the expense of integrating these technologies into a production-ready vehicle, the price tag skyrockets far beyond what most consumers could reasonably afford. Until the cost of these critical parts drops substantially, mass-market adoption of self-driving cars will remain out of reach.

Robotaxis in the U.S. Today

While you can’t buy a fully self-driving car yet, you can ride in one in a few American cities. Waymo currently operates driverless taxis in Phoenix, San Francisco, Los Angeles, Atlanta, and Austin. The rides are hailed via app, with no driver behind the wheel.

Tesla has also begun a limited pilot in Austin with a few vehicles. These robotaxi rides, though, still require human supervisors and are considered closer to Level 2–3 than Level 4.

These projects demonstrate how Level 4 is gradually entering public life, even if Level 5 remains a distant dream.