From Luxury to Liability: Legal Blind Spots in AI-Related Vehicle Collisions

By: Natasha Agarwal
Natasha Agarwal is a freshman studying Law & Society at Penn’s College of Arts and Sciences.

In 2023, a self-driving car at Cruise robotaxi company dragged a jaywalking pedestrian across a bustling San Francisco street [1]. The incident quickly went viral—sparking public outrage, legal confusion, and a corporate scandal. Yet it also revealed a broader issue: despite the rising prevalence of AI-powered vehicles, collision laws still take a back seat. 

By 2030, over 60% of vehicles sold are projected to feature some level of automation, from fully driverless models to human-operated cars equipped with lane assist or sensor technologies [2]. As these vehicles become increasingly popular among individuals, rideshare services, and businesses, establishing clear liability standards for collisions involving automated vehicles  becomes essential to ensure public trust, accountability, and consumer protection. Yet assigning fault remains a gray area. Automation varies across vehicle companies and states, blurring the lines between human and machine responsibility while driving.

Automated vehicles operate on a spectrum, classified from fully manual (Level 0) to fully autonomous (Level 5) [3]. Collisions of vehicles at the middle or higher levels—where human oversight is still recommended—often spark debates over who is responsible: car owners, software, or manufacturers. Not only does this problem complicate court proceedings, but it highlights the growing gap between innovation and the law’s ability to keep up.

Courts currently approach each AV collision as an independent case, leading to delayed compensation and inconsistent judgments. The 2023 Cruise incident demonstrates this issue: the company initially claimed the human driver who struck the pedestrian was solely responsible, while officials and civilians contended that Cruise’s AI system misperformed. The resulting controversy, coupled with allegations of a corporate cover-up, damaged the company’s reputation to the point where General Motors, Cruise’s parent company, shut down the division entirely [4]. More broadly, this incident suggests a defensive industry posture that prioritizes reputation over safety [5]. Such practices undermine public trust and raise serious ethical questions surrounding accountability and transparency. 

Despite sophisticated sensors and systems, automated systems are largely fallible. Pre-release safety testing remains largely self-governed, and as of 2024, only twelve U.S. states legally mandate testing or piloting of AVs—with procedures varying widely in magnitude and enforcement [6]. Federal frameworks were written for human-driven vehicles and do not account for autonomous decision making, sensor calibration, or live visual processing [7]. AI cannot yet make impulsive decisions akin to human judgement as training is still an evolving science.
Without standard guidelines or oversight, companies are left to police their own technologies, leaving room for safety errors to occur  before vehicles reach the market.

Legislators can address these risks by introducing nationalized safety benchmarks like blackbox testing—a level one to five vehicle-simulation method that enables AVs to master various conditions [8]. This testing method relies on optimization, reinforcement, and other human psychological strategies to evaluate software without revealing the machine’s internal codes. A phased rollout—starting with the testing of higher-automation vehicles (Levels 3-5)—can safeguard public welfare while allowing room for continued innovation. Collecting and publishing performance data before vehicles are released into the market will help regulators, insurers, and owners understand how these systems behave in stressful situations. 

Level one to four vehicle collisions require more situational fault assignments, depending on the driver’s mental state and ability to intervene during an adverse situation. National guidelines should explicitly account for the four driver scenarios that liability expert Jeffrey Gurney identifies—the distracted, diminished capabilities, disabled, or attentive driver [9]. This way, collision responsibility is proportionally and contextually assigned among humans, sellers, and manufacturers. Determining these fault assignments at the national level would provide courts with a consistent framework for liability evaluation, reduce payout delays, and curb corporate cover-ups. Nevertheless, such efforts may face resistance from states hoping to control traffic and vehicle regulations. Collaboration between federal and state officials can occur through joint legislative committees and lobbying initiatives that include insights from DMVs, transportation agencies, automakers, and insurance firms. Providing federal incentives like funding for pilot programs, technical testing, or access to collision history can encourage state participation—allowing them to maintain autonomy while adopting improved frameworks.

Assigning fault after automated vehicle collisions can take months or years, and victims often encounter compensation delays. Insurance policies should be updated to reflect automation-related risks, such as system failures or cybersecurity breaches. Collecting data on algorithm error rates would help these companies assign limits proportional to each vehicle automation level. For example, if Level 3 vehicles collide less than Level 2 ones, insurance for Level 3 vehicles may cost less. Tiered limits more accurately reflect the shared responsibility among drivers, sellers, and manufacturers while promoting timely compensation for victims. Safety reports and proportional coverage also improve clarity and reduce blame-related tension.

While these reforms require significant coordination across state, federal, and corporate bodies, they are feasible through gradual implementation. The National Highway Traffic Safety Administration could oversee standardized safety protocols for automated vehicle testing, while the Department of Transportation and state Departments of Motor Vehicles could manage the implementation of national frameworks in collision reporting, proportional fault assignment, and victim compensation. Although full adoption of these solutions will take several years, starting with higher-automation (Level 3–5) vehicles will allow for scalable and timely reform. With these measures in place, AI will continue to develop as a safe, transparent, and collaborative force in the transportation industry—ensuring that the law is no longer the blind spot on our roads.

Bibliography
[1] https://doi.org/10.48550/arXiv.2406.05281
[2]https://www.iea.org/reports/by-2030-evs-represent-more-than-60-of-vehicles-sold-globally-and-require-an-adequate-surge-in-chargers-installed-in-buildings 
[3] https://www.epa.gov/greenvehicles/self-driving-vehicles 
[4] https://doi.org/10.48550/arXiv.2406.05281
[5] https://www.brookings.edu/articles/setting-the-standard-of-liability-for-self-driving-cars/ 
[6]https://www.bakerdonelson.com/autonomous-vehicle-statutes-and-regulations-across-the-50-states 
[7] https://www.nhtsa.gov/press-releases/av-framework-plan-modernize-safety-standards 
[8] https://doi.org/10.1613/jair.1.12716
[9]  https://ssrn.com/abstract=2352108