2–Automated system can actually conduct some parts of driving while human monitors the environment and performs the rest of the driving. System can provide continuous assistance with both steering AND acceleration/braking.

3–Automated system can actually conduct some of the driving and monitor the environment in some instances, but human must be ready to take back control when the system indicates. Driver can take their hands off the wheel and feet off the pedals. Known as conditional automation.

4–Automated system can drive and monitor the environment, and human does not need to take back control, but the vehicle can operate only in certain environments and conditions. For example, vehicles must operate only within a certain geo-fenced area.

5–Automated system can perform all driving tasks under all conditions that a human could perform.

Many of the lower levels of autonomy rely on Advanced Driver Assistance Systems (ADAS) for the technology to enable vehicles to behave semi-autonomously. Systems such as collision detection, blind-spot warnings, lane-changing assistance, automatic emergency braking and adaptive cruise control use sensors and cameras to assist with driving without taking over the task completely. ADAS systems allow the vehicle to perform certain tasks with the human monitoring the vehicle.

ADS, or Automated Driving System, is a step above ADAS, and vehicles contain both the hardware and software capable of driving the vehicle without human monitoring or intervention. With ADS systems, the vehicle can handle all driving tasks on its own, allowing the human passenger to engage in other activities.

A primary concern for fully autonomous vehicles is decision-making in no-win situations. The vehicle must be programmed to determine whether to protect its passenger at the risk of hitting another vehicle or pedestrian, or to prioritize the safety of others over its passenger.

Availability

Currently, no level 4 or 5 vehicles are available for sale to the general public in the United States. Vehicles with Level 2 and Level 3 features offer partial autonomy, but the driver must remain alert and ready to take control. These systems are often limited to specific conditions, such as highways in clear weather.

BMW offers level 2 and level 3 autonomous features on its 7 Series. The Level 2 Highway Assistant allows speed, steering, and lane control with the driver taking their hands off the wheel for longer periods as long as they are paying attention to the vehicle and surroundings. It allows hands-free driving up to 85 mph when conditions allow as well as automatic lane changes activated by the driver looking in the corresponding exterior side mirror. The driver is always responsible for the actions of the vehicle. The Active Driving Assistant Pro includes Steering Assistant, Lane Change Assistant and Distance Control, Emergency Stop Assistant and Evasion Assistant to aid the driver during the sudden presence of other vehicles or pedestrians in front of the vehicle. The Level 3 Personal Pilot, available only in Germany, allows drivers to divert their attention from the road and take their hands off the wheel on certain highways at speeds up to 37 mph, though they must be prepared to resume control when prompted.

Mercedes-Benz offers its Level 3 Drive Pilot system on a subscription basis for EQS and S-Class vehicles in California and Nevada. Honda's Sensing Elite system, different from the Honda Sensing, is available on models in Japan.

Many manufacturers offer level 2 technology – Tesla’s Autopilot, Ford’s BlueCruise, Volvo’s Pilot Assist, Cadillac’s Super Cruise, and GM’s Super Cruise are some of the more well-known examples. Kia, Nissan, Hyundai, Volkswagen, Honda, Lincoln, Volvo, and Audi also offer level 2 options.

Autonomous Taxis

While personally owned autonomous vehicles remain in development, autonomous ride-hailing services are operational in limited areas. Waymo One launched the first fully autonomous ride-hailing service in Phoenix in 2018 and became available to the public in metro Phoenix in 2020.

In 2022, the program expanded to San Francisco and the Phoenix airport. In 2023, it expanded service to Scottsdale and doubled availability in Phoenix, and began testing in LA. In 2024, it became available in both LA and San Francisco. The company plans to enter the Dallas market in 2026, partnering with Avis Budget Group for fleet management. Currently, Waymo provides more than 250,000 trips a week with roughly 1,500 vehicles. Waymo has surpassed 10 million rides and operates in LA, Phoenix, San Francisco, Atlanta, and Austin and will operate in Washington, D.C. and Miami next year.

Waymo was approved in September to test autonomous vehicles in Manhattan and downtown Brooklyn, testing 8 autonomous vehicles until the end of September. During this testing a safety driver is required to be behind the wheel when the vehicles are in operation. Waymo has also just been approved to start testing at the San Francisco airport.

As of July 8, teens aged 14-17 can travel independently with Waymo when linked to a parent or guardian’s account, allowing teens to hail a fully autonomous ride within the available area. Trained Rider Support agents are available to assist teens during a ride, and parents can be looped in if needed. Parents receive all ride receipts and teens can share the status of the ride with their parents for real-time updates.

Tesla is planning to start offering a public car service in San Francisco. While the company refers to it as “robotaxis”, because of regulatory issues, the vehicles will have a human driver in the vehicle. The company plans to first offer the service to friends and family of employees and select members of the public.

Currently, Tesla has permission from the California Department of Motor Vehicles to test driverless vehicles with a driver behind the wheel. In June in Austin, Tesla began inviting members of the public to ride in robotaxis with employees in the passenger front seat in order to intervene if necessary. Tesla is seeking approval to operate its robotaxis in California, Nevada, Florida, and Arizona.

Uber has partnered with Lucid and Nuro to begin providing an autonomous vehicle service. With Uber’s global platform combined with the Nuro Level 4 autonomous systems and Lucid electric vehicles, the company plans to launch in a yet undesignated city in the United States in 2026. The company plans to launch 20,000 or more vehicles equipped with autonomous technology over the next six years, making it one of the largest autonomous vehicle services so far. However, the companies need to meet state regulatory requirements before the vehicles will be found on city streets.

Lyft has also announced plans to introduce autonomous shuttles in certain cities and airports in late 2026. The vehicles, manufactured by Benteler’s Holon brand, will have no steering wheels or pedals and will accommodate up to 15 passengers, with nine inward facing seats and six standing.

Zoox just launched September 10 in Las Vegas with an electric robotaxi with no steering wheel or pedals. The vehicles have floor to ceiling windows, and carry 4 passengers. They didn’t retrofit an existing vehicle, they created a new vehicle designed to carry passengers only.

Safety

The CDC reports that motor vehicles are the leading cause of death in the U.S., and the IIHS reports that as of 2023 there were 40,901 motor vehicle fatalities. Most accidents are the result of human error, be it distracted driving, speeding, driving while impaired, and other behaviors.

Autonomous vehicles are touted as being safer than human drivers. However, a number of accidents have been reported where the autonomous technology didn’t behave exactly as programmed or ran into other problems. Companies with autonomous vehicles on the road are required to report accidents to the National Highway Traffic Safety Administration (NHTSA).

NHTSA and the U.S. Department of Transportation issued the Federal Automated Vehicles Policy in 2016 to provide guidance, ensure public safety, and start to establish a base of rules surrounding autonomous vehicles. Since 2016, versions 2.0, 3.0, and 4.0 have been further developed as autonomous technologies have advanced and changed. AV 4.0 is designed around principles of AV technology, administration efforts to support AV technology and advancement, and opportunities for collaboration between government agencies and automated technology innovators.

In 2021, NHTSA issued a Standing General Order to vehicle manufacturers and operators to report certain crashes involving vehicles equipped with automated driving systems or Level 2 advanced driver assistance systems. This allows NHTSA to investigate crashes that raise safety concerns. The order was revised in 2021, 2023, and 2025.

A total of 3,979 accidents were reported between 2019 and 2024, and this includes vehicles with ADS and ADAS systems. While the number of incidents is increasing as more of these vehicles enter service, their accident rate remains comparatively low. Waymo reports its vehicles have 92 percent fewer crashes with pedestrian injuries, 82 percent fewer injuries to cyclists and motorcyclists, and 96 percent fewer accidents at intersections with other vehicles. Accidents at intersections with other vehicles is a leading cause of injury to drivers.

The bottom line is that fully autonomous vehicles - those at Level 5 - are not yet available for public sale. As with any emerging technology, there are issues that arise along the development chain requiring continual testing and training of the technology. Once vehicles become fully autonomous and safety reliably is no longer a concern, then the accident rate could start to decline. Autonomous taxis are available in certain limited areas such as Phoenix and San Francisco, areas where snow is not a concern and the weather is generally favorable.

Public perception is important, and currently, people are wary of the technology being able to reliably drive like a human in all conditions, and be able to respond timely and appropriately. A 2022 study by the Pew Research Center showed that 44 percent of Americans think the widespread use of autonomous vehicles would be a bad idea, 29 percent were unsure, and 26 percent thought they were a good idea. Study participants were more wary of riding in an autonomous vehicle, with 63 percent saying they would turn down a ride in one. As far as sharing the road with such vehicles, 45 percent expressed discomfort with the idea of sharing the road with autonomous vehicles, and 18 percent would not be comfortable at all with such vehicles on the road.

Regulations

One of the biggest issues with autonomous vehicles is liability; if a vehicle without a driver causes an accident, who is responsible for damages or injuries? Is it the owner of the vehicle, who had nothing to do with the design of the driving system and was in the back seat of the vehicle? Is it the manufacturer who developed and built the vehicle with the autonomous technology? Whose insurance policy is activated when a driverless vehicle has an accident?

Drivers must have passed a driving test and maintain a valid driver's license in order to operate vehicles on the road. In some states, the minimum driving age is 15. With an autonomous vehicle, what assurance that the vehicle can accurately navigate the streets, highways, side roads and parking lots safely without a driver present is needed? Will that autonomous vehicle perform the same in other states and in differing weather conditions? What level of testing is required? How do states authorize a fully autonomous vehicle for sale to the public for use in navigating all driving areas in the country and in all road conditions? These are just a few pressing and important questions that come to mind.

As fully autonomous vehicles are not yet available for purchase by the public, states are working to determine what regulations are needed to ensure the public’s safety. States continue to develop regulations as the technology continues to evolve. The IIHS has a chart with current regulations which can be found in Autonomous Vehicle Driving Laws by State.

The Autonomous Vehicle Accessibility Act, which was introduced in 2024, was reintroduced on July 22, 2025, by Representatives Greg Stanton and Brian Mast to help people with disabilities access the mobility benefits of ride-hail autonomous vehicle services. The bill ensures that the passengers are legally protected if the vehicle is pulled over or other issues occur. Such passengers may not have a driver’s license due to their disability. The act also requires the Department of Transportation to study best practices for modification of public transportation infrastructure to better serve those with disabilities and help them find, access, and use ride-hail autonomous vehicles.

Autonomous Trucks/Platooning

While autonomous vehicles are still being developed for the general public, trucks are a different matter. The commercial trucking industry is turning to automation to address a driver shortage, which the American Trucking Associations estimates at 60,000 drivers. Gatik has more than 100 autonomous trucks operating in Texas, Arkansas and Canada, partnering with companies like Walmart, Kroger, Tyson and others to transport goods within the supply chains. Trucks run on fixed routes, highways, and are more structured that private drivers.

Aurora Innovation Inc. has started running autonomous trucks day and night between Dallas and Houston, and has a terminal in Phoenix. Adding the nighttime trips doubles Aurora’s capacity, shortening delivery times. The company provides a livestream showing the trucks on the road so stakeholders can see how the vehicles perform on the road. The vehicles can detect objects more than 450 meters away and identify hazards 11 seconds sooner than a human driver. Unlike human drivers, who are limited to 11-hour shifts, autonomous trucks can operate continuously. Since May, the company has logged more than 20,000 autonomous miles, although the company has humans behind the wheel as backup. The company hopes to have traveling in the rain mastered by the end of the year.

Kodiak, founded in 2018, provides autonomous trucking services coast-to-coast.

In 2021, Kodiak completed its first driverless test on a test track, then in 2022 began its first coast-to-coast commercial delivery, with service from Dallas to Atlanta. In 2023, the freight company Loadsmith ordered 800 Kodiak Driver systems to launch a trucking operation dedicated to autonomous freight.

Truck platooning is a little different. A driver operates the lead truck, while the second truck is connected via encrypted communications and follows the lead truck’s steering, acceleration and braking. The trucks can be as close as 30 meters, closer than the safe distance for human drivers due to reaction time. A platooning system is being tested along I-70 between Ohio and Indiana delivering shipments for EASE Logistics. The trucks have covered 2,700 miles, made nearly 50 deliveries, and operated for almost 50 hours since the program began in April. The benefits are similar to other autonomous technology as only one driver is needed instead of two, safety is improved, and EASE has already seen increased fuel efficiency for the second truck following the lead truck.

Pros/Cons

As mentioned earlier, autonomous vehicles that obey all traffic regulations, don’t get distracted or sleepy, and don’t engage in other activities while driving are promised to reduce accidents on the road. Passengers can engage in other activities or rest along the way, traffic will travel more efficiently, and many parking issues would be alleviated as the vehicle can go on to the next destination or drop off the passenger and then find a place to park. The availability of autonomous vehicles also opens transportation up to the disabled community, making it easier for them to obtain rides to the doctor or store without having to rely on others for transportation.

However it’s not all roses; until autonomous driving has been perfected, there will be accidents and potential injuries. Driving requires a tremendous amount of data and vehicles will have to collect and store it, and changes in roads will need to be uploaded to the vehicle. Hacking of the vehicle’s software is a concern, as is data protection. As mentioned earlier, liability is a huge concern, since with no one driving the vehicle, who is responsible for an accident caused by the vehicle? Would it be the owner, the manufacturer, the software engineer, who? Lawsuits have been filed against Tesla for actions of vehicles in Autopilot mode which is not fully autonomous, and other manufacturers have been sued for accidents that occurred while autonomous vehicles were being tested.

Cases

With early accidents, there are bound to be lawsuits. One of the problems with semi-autonomous vehicles is that their true capabilities are misunderstood, and buyers assume the vehicles can do more than they actually can.

In the case In re Gen. Motors Co. Sec. Litig., 773 F. Supp. 3d 429 (E.D. Mich. 2025), investors who bought GM securities between February 24, 2021 and November 8, 2023 (the class period) filed a putative class action for securities fraud, alleging that GM, Cruise LLC, and certain executives misled the market about the capabilities of Cruise’s autonomous vehicles.

During the class period, executives described Cruise as “fully driverless,” “truly driverless,” and “Level 4,” and touted the safety advantages of Cruise AVs over human drivers. The allegations intensified after an October 2, 2023 incident in which a Cruise AV struck a pedestrian, pinned her underneath, then began moving again and dragged her about 20 feet.

Plaintiffs say GM and Cruise issued misleading post-accident statements and videos, and that three subsequent corrective disclosures (Oct. 24, Oct. 26, Nov. 8) triggered stock drops and investor losses.

The defendants filed a motion to dismiss the suit for failure to state a claim. In its analysis, the court grouped the challenged statements into pre-accident “Technology Statements” (commercialization, “fully driverless,” “Level 4,” permit, safety comparisons) and post-accident “Accident Statements.” Defendants contested falsity, scienter (knowledge or reckless disregard for whether statements were false), and loss causation for most categories.

The court granted the motion in part and denied it in part. Claims tied to the commercialization statements could proceed against all defendants; claims based on the “Technology Statements” could proceed against Cruise, its CEO, and its COO; and the “Accident Statements” claims could proceed against Cruise and its CEO. The court granted the motion as to all other challenged statements.

The case Benavides v. Tesla, Inc., 2025 U.S. Dist. LEXIS 121472 (S.D. Fla. 2025) arises from an April 25, 2019 crash in which a Tesla operating with driver-assistance features struck a parked vehicle, killing one person and seriously injuring another. The driver admitted he was looking down to retrieve his phone at the time of the accident. He testified that he understood Tesla’s Autopilot was not fully self-driving, but he had also expected copilot-like assistance from features such as lane-keeping, obstacle detection, and crash avoidance. The record showed that the driver had experienced 23 “strikeouts” prior to the accident. Each “strikeout” meant the Tesla’s Autopilot features were disabled for the remainder of that particular driving trip, which occurred after the driver received three warnings for inattentiveness. However, the driver skirted each strikeout by pulling the vehicle over and parking it for a moment, thereby resetting the warning counter, before resuming operation of the vehicle.

Plaintiffs, who were representatives of the estate of the decedent, asserted product-liability claims; Tesla sought to exclude key expert opinions and win summary judgment, arguing driver negligence was the sole proximate cause and that its design and warnings were adequate. Under the Daubert standard, the court ruled that much of plaintiffs’ technical case was admissible as evidence.

An expert for the plaintiffs, Alan Moore, had sufficiently reliable foundations to offer his opinions on defects in Tesla’s driver monitoring system (DMS), failure to restrict use to the operational design domain (ODD), and the failure to warn the driver or apply the brakes. However, his opinions on inadequate training and “beta software” risks were excluded. The court found the opinions of Dr. Mary Cummings, another expert for the plaintiffs, had a sufficiently reliable basis for her opinions on the use of Autopilot outside its ODD, the unreliability of steering wheel torque as a measure of driver engagement, the lack of warning or brake engagement from the obstacle detection feature, and the inadequacy of both Tesla’s warnings and its training. On plaintiffs’ challenge to defense expert Ryan Harrington, his testimony regarding Phase 1 testing, which occurred during the day at a much slower speed, was excluded for lack of substantial similarity. However, his testimony on Phase 2 testing, which examined higher speeds at night, and Phase 3 testing, which measured the automatic brakes and forward collision warning at night, was admissible; any differences between the test conditions and the conditions at the time of the accident would go to the weight of the evidence. The opinions of Tesla expert Dr. Bernard F. Pettingill were excluded in full.

On summary judgment, the court dismissed the manufacturing defect and negligent misrepresentation claims, but allowed the strict-liability design defect and failure-to-warn claims to proceed, as well as plaintiffs’ petition for punitive damages. The court found genuine disputes of material fact over whether Tesla’s design, including the ODD limits and DMS, was defective and whether warnings, particularly in-vehicle messages and an electronic owner’s manual, were adequate and accessible. The judges rejected Tesla’s sole-proximate-cause argument at this stage, holding that a jury could find Tesla’s design and warnings were a substantial factor and that driver negligence was foreseeable, leaving comparative-fault issues for trial. The jury awarded the plaintiffs $243 million, assigning 67% of the fault to the driver and 33% to Tesla. As of September 2nd, Tesla has asked a federal court to dismiss the damages based on claims that the jury was misled by the bringing up of Elon Musk during the trial.

These cases highlight the difference in the public perception of what the vehicles are capable of based on advertising and names of features, and the actual capabilities of the vehicles. Until vehicles are truly level 5 autonomous, mores cases like these are bound to arise.

Conclusion 

Fully autonomous vehicles that can drive completely without human intervention are still a ways off, even with the few autonomous taxis that are available currently. Regulations are needed to establish liability when an accident, however unlikely, occurs.

Assuming that autonomous vehicles become available to the public and the fleet of American vehicles becomes more autonomous, insurance is bound to change as well. Will people need coverage other than for hail, vandalism, and falling trees? Will liability coverage really be necessary if vehicles can reliably drive without error? Will people give up owning vehicles in order to travel instead by oncall autonomous vehicle services?

Predictions as to when autonomous vehicles will be available have missed the mark; however as the industry progresses, autonomous vehicles look like they will be an option in the future. What happens then within the realms of transportation, insurance, and regulation will be interesting to see.