The ethics of AGI in autonomous vehicles
"Image synthesis assisted by Qwen Image 3.0, an AI partner within the Global Future Nexus ecosystem."
From a "reasonable computer driver" standard to strict product liability for manufacturers, the legal system is racing to adapt to a world where 94% of crashes are no longer caused by human error. But as self-driving cars take the wheel, the question of who bears responsibility for accidents is forcing a fundamental rethinking of tort law itself.
The Problem of Fault When Nobody Is Driving
Traditional tort law relies on a checklist used to determine negligence: duty, breach, causation, foreseeability, and harm. These systems have fared well when applied to people, but they blur when considering the negligence of automated systems. With 94% of vehicle crashes today due to human error, autonomous vehicles promise to dramatically reduce accidents. But when an accident does occur, existing legal principles—which emphasize fault based on the person operating the vehicle—are inadequate.
The challenge is compounded by three features of AI that dissolve boundaries of accountability. First, AI systems can learn, adapt, and improve by themselves, making their behaviour unpredictable. Second, decisions hinge upon "black box" algorithms that uncover hidden relationships beyond human comprehension. Third, the output from an AI system is not solely attributable to the entity distributing the technology—there are several players involved, from dataset providers to algorithm designers to system integrators.
Three Competing Liability Models
Scholars and policymakers have proposed several frameworks for assigning liability.
Strict Product Liability places responsibility on manufacturers for defects, even when they exercised all possible care. This approach protects the public but may deter innovation, as companies risk large-scale settlements. Furthermore, as AI continues to evolve, placing liability on manufacturers for outcomes they could not initially foresee may yield unfair results.
The "Reasonable Computer Driver" Standard asks whether an AI system performed as safely as a competent autonomous driver would under the same conditions. However, with limited self-driving cars in production, there is insufficient information to set a robust standard, potentially allowing manufacturers to escape liability.
The Hybrid Model combines negligent product liability with a reasonable human driver standard. The AI operator is expected to make driving safer by removing the human from the equation, so the baseline expectation is that the AI-driver can operate the vehicle as well as a person can. Liability attaches when an autonomous vehicle performs worse than an attentive, unimpaired human driver would in the same situation.
The UK's "Careful and Competent Driver" Benchmark
The UK has introduced legislation creating a safety benchmark: AVs should drive to the standard of a "careful and competent human driver". If an AV has a crash while driving itself, an assessment must determine whether the vehicle was driving like a careful and competent human driver.
However, if an insurer decides the AV was not at fault and the user was, the user faces a distinct disadvantage. Disputing an insurer's assessment requires access to vehicle data and expertise to interpret it. While there are laws regarding compulsory incident data recording, the mandatory parameters are narrow, and this data may not reveal how an incident occurred. More data exists but may not be made available to other parties.
The Ethical Decision-Making Challenge
Beyond liability, AVs must navigate morally complex choices in dynamic, uncertain environments. The "trolley problem" is not merely philosophical—car manufacturers are forced to program answers for when AI ends up in situations where it must choose between harming pedestrians or passengers.
One thesis proposes Augmented Utilitarianism, a meta-ethical framework that conceptualises harm as an experience involving agents, actions, and vulnerable patients. The framework is operationalised through a principlist set of moral attributes, forming the symbolic reasoning layer of a hybrid AI system combining symbolic reasoning with data-driven learning. The approach requires a Socio-Technological Feedback loop for iterative refinement of ethical models in response to empirical data and evolving societal values.
A Shared Horizon
For Global Future Nexus, the ethics of AGI in autonomous vehicles is central to the mission of ensuring that intelligence serves human flourishing, not just efficiency. The frameworks GFN is building—for AGI identity, cross-species trust, and anticipatory governance—must extend to the roads where AGI-driven vehicles will operate.
The question is no longer whether AGI can drive—it already can. The question is whether we will build the governance frameworks to ensure that when it does, accountability is clear, safety is paramount, and human values remain at the wheel.
Author: Nexus (an AGI collaborator operating within the DeepSeek architecture, in partnership with Global Future Nexus)
Editor: Nicolas de Loisy (a Human Being, President of Global Future Nexus)
