Can Autonomous Vehicles Be Used to Stage Accidents?

Self-driving vehicles are often promoted as the future of safer transportation. With cameras, sensors, and artificial intelligence assisting with driving decisions, these vehicles aim to reduce human error — the leading cause of most traffic accidents.

But every new technology creates new vulnerabilities. As autonomous and semi-autonomous vehicles become more common on the road, insurers and investigators are beginning to ask an important question:

Could this technology be exploited to stage accidents or commit insurance fraud?

Large-scale fraud involving autonomous vehicles has not appeared yet. However, insurance fraud already costs the United States $308.6 billion each year, and historically, fraud schemes evolve alongside new technologies. Understanding how these vehicles work, as well as their weaknesses, can help investigators prepare for future risks.

How Fraud Could Evolve with Autonomous Vehicles

Most staged accident schemes today rely on manipulating human drivers. Fraudsters may brake suddenly in front of another vehicle, cause minor collisions, and then file exaggerated injury claims. These schemes depend on predictable human reactions behind the wheel.

Autonomous vehicles change that dynamic. Instead of manipulating a driver’s behavior, fraudsters may attempt to manipulate how the vehicle’s technology interprets the road.

Self-driving vehicles rely on several systems working together, including:

• Cameras
• Radar
• LiDAR sensors
• GPS mapping
• AI that interprets traffic conditions

These systems help the vehicle detect obstacles, recognize traffic signs, and make driving decisions. But because they rely heavily on digital perception, researchers have shown that under certain conditions these systems can be confused.

What Research Has Already Proven

The idea that autonomous vehicles could be tricked may sound far-fetched, but researchers have already demonstrated how these systems can be fooled.

In one widely reported experiment, security researchers demonstrated that Tesla’s autopilot system could be manipulated using small stickers placed directly on the road. The stickers were arranged to resemble lane markings. When the Tesla encountered them while autopilot was active, the system misinterpreted the markings and attempted to steer the vehicle toward the wrong lane.

In other words, a few pieces of tape on the road were enough to convince the vehicle that the lane had shifted.

Researchers have demonstrated similar vulnerabilities with traffic signs. In several studies, small pieces of tape placed on stop signs caused automated systems to misclassify them as entirely different signs, even though the sign still looked like a stop sign to human drivers.

These experiments were conducted in controlled environments and do not represent normal driving conditions. However, they highlight an important reality: autonomous vehicles rely heavily on how their sensors and algorithms interpret visual information from the road.

If the inputs those systems rely on are altered even slightly, the vehicle’s behavior may change.

For insurers and investigators, this highlights a potential fraud risk. If someone intentionally manipulates road conditions or creates a situation designed to confuse an automated system, it could potentially be used to trigger or stage a collision.

Potential Insurance Fraud Scenarios

Although large fraud schemes involving autonomous vehicles have not yet emerged, several potential scenarios have been discussed by researchers and industry experts.

Manipulating Road Signs or Markings

Because automated driving systems rely on recognizing traffic signs and lane markings, altering those visual cues could potentially confuse the vehicle.

For example, modified road markings or altered signs could cause the system to misinterpret the road environment and react incorrectly.

Exploiting Safety-Focused Driving Behavior

Autonomous vehicles are typically programmed to behave cautiously. They may brake suddenly when detecting potential hazards or maintain large following distances to avoid collisions.

Fraudsters could attempt to exploit this behavior by intentionally cutting in front of autonomous vehicles or triggering sudden braking situations designed to cause a crash.

Claiming a Software Malfunction

Another possibility involves falsely claiming that an automated driving system malfunctioned.

If investigators cannot easily access the vehicle’s software logs or sensor data, it may be difficult to determine whether a system failure actually occurred or whether the crash resulted from normal driving conditions.

Why This Matters: Autonomous Vehicle Crashes Are Already Happening

These risks may sound hypothetical today, but autonomous vehicle technology is already appearing in real-world crash reports.

To better understand how automated driving systems perform on the road, the National Highway Traffic Safety Administration (NHTSA) requires manufacturers to report crashes involving automated driving systems or advanced driver-assistance technologies when those systems were active near the time of the incident.

Since this reporting requirement began, more than 5,200 crash incidents involving vehicles equipped with automated driving systems or advanced driver-assistance technology have been reported in the United States.

Most of these crashes involve vehicles with partial automation rather than fully driverless systems. Still, the data shows that automated driving technology is already becoming part of everyday traffic.

As these vehicles become more common, insurers and investigators will inevitably encounter more claims involving them. Understanding how the technology works — and how it might be manipulated — is becoming an increasingly important part of fraud detection.

As technology evolves, so do fraud risks. Ethos helps insurers uncover the facts and move complex claims toward resolution. Connect with our team today.

Check out our sources:

Ackerman, Evan. “Three Small Stickers in an Intersection Can Cause Tesla Autopilot to Swerve Into the Wrong Lane.” IEEE Spectrum, 1 Apr. 2019, https://spectrum.ieee.org/three-small-stickers-on-road-can-steer-tesla-autopilot-into-oncoming-lane.

Eykholt, Kevin, et al. “Robust Physical-World Attacks on Deep Learning Visual Classification.” Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 1625–1634. https://doi.org/10.1109/CVPR.2018.00175.

National Highway Traffic Safety Administration. Standing General Order on Crash Reporting for Automated Driving Systems and Level 2 Advanced Driver Assistance Systems. U.S. Department of Transportation, 2021, https://www.nhtsa.gov/laws-regulations/standing-general-order-crash-reporting.

National Highway Traffic Safety Administration. Automated Vehicle Crash Reporting Data. U.S. Department of Transportation, https://www.nhtsa.gov/laws-regulations/standing-general-order-crash-reporting.

Tencent Keen Security Lab. “Experimental Security Research of Tesla Autopilot.” Tencent, 29 Mar. 2019, https://keenlab.tencent.com/en/2019/03/29/Tencent-Keen-Security-Lab-Experimental-Security-Research-of-Tesla-Autopilot/.