Algorithm at the Wheel: From Negligence to Algorithmic Liability in the Age of Tesla Autopilot

In the relentless march of time, when innovation becomes subservient to anxiety, and the value of human life is measured against technologically mediated standards, an inevitable question arises: who bears responsibility for the control and consequences of driverless vehicles? Frederick Winslow Taylor’s efficiency-driven concept of a ‘machine-like human’ has now transformed into a reality where machines replace humans. When monitoring and evaluation along with control , instruction, knowledge, and consequence ; all these layers are vested in artificial intelligence, the traditional basis of determining liability is naturally questioned.

When monitoring and evaluation, along with the control, command, cognition, and consequence (4C’s)are vested in artificial intelligence, the traditional basis of determining liability is naturally questioned.

Artificial Intelligence (AI) is a computational system that learns from data and generates autonomous decisions or actions through inference.

Although the concept of autonomous vehicles dates back to the 1920s, its practical development emerged in the 1980s through the Navlab and Prometheus projects. Subsequently, companies such as Tesla, Google, Volvo, and Audi have contributed to advancing this technology. In particular, in 2014, Elon Musk introduced Autopilot in the Tesla Model S, which elevated autonomous technology by enabling partial control over steering, braking, and speed. At present, Level 2 automation is prevalent, where human supervision remains essential. At the same time, Tesla has expanded its influence through fast charging facilities (approximately 30 minutes) and through models like the Tesla Model 3, one of the world’s best-selling plug-in electric cars. The company aims to advance sustainable electric mobility through affordable and advanced vehicles. According to Morgan Stanley, autonomous vehicles could reduce fuel costs by approximately 488 billion dollars annually, further highlighting their economic significance.

On one hand, autonomous vehicles represent new engineering challenges that are gradually being resolved. On the other hand, social and ethical issues are often discussed in terms of the trolley problem, which is inherently misleading. In the context of autonomous vehicles, one of the most complex questions is this ethical dilemma: where a vehicle, facing an unavoidable accident, must choose between two harms—striking pedestrians ahead or diverting and causing the passenger’s death. This situation is not merely technical; it raises profound legal and ethical questions: whom should the algorithm prioritize, and upon whom should the responsibility for that decision fall? It is at this point that autonomous technology challenges conventional notions of liability.

The development of autonomous vehicles began to expand in the global market following the enactment of the ISTEA Act in the United States in 1991, growing rapidly like a blooming field. However, in 2016, a fatal accident in Florida involving a Tesla vehicle operating under Autopilot brought the legal debate surrounding autonomous technology into sharp focus for the first time. It raised a critical question: does liability lie with the driver, or with the technology?

There is a fundamental structural difference between human and computer decision-making. Humans make decisions through a combination of perception, cognition, and experiential learning. In contrast, computers transform sensor-derived input data into decision outputs through algorithmic processing, including recognition and computation, based on predefined models. In autonomous vehicles, this process relies heavily on laser-based ranging to create high-precision three-dimensional environmental maps and to detect objects such as roads, vehicles, pedestrians, and obstacles. However, real-world uncertainty, incomplete data, and sensor limitations complicate this decision-making process.

In analysing the ethical dilemmas of autonomous vehicles, two major theories are often applied: Utilitarianism and Deontology. Under Utilitarianism, the morality of a decision depends on its consequences, with the goal of achieving the greatest good for the greatest number. For example, in a brake-failure scenario, diverting the vehicle to sacrifice one person in order to save five may be considered acceptable. However, this approach may disregard individual rights. In contrast, Deontology emphasizes duty and moral rules. It holds that intentionally harming an innocent person is unacceptable, even if it leads to a greater good. Therefore, in the same situation, the vehicle may continue on its path, as actively causing harm is considered morally impermissible.

Under tort law, liability is not based solely on actual damage but on the violation of a legal right, as established by the theories of Salmond and Winfield, since “damage without violation of a legal right is not actionable.” In this context, damnum sine injuria signifies that even where there is economic loss or inconvenience, no claim arises if no legal right is infringed. For instance, in Acton v Blundell (1843), no liability arose from the lawful use of property; in Dickson v Reuter’s Telegraph Co. (1877), the action failed due to the absence of duty; and in Day v Browning (1878), mere inconvenience was not treated as legal injury. Similarly, in self-driving vehicles, harm arising solely from algorithmic decisions or mere technological malfunction is generally non-actionable. In contrast, injuria sine damno establishes that the violation of a legal right is itself actionable even without actual damage, as recognized in Ashby v White (1703) and Bhim Singh v State of J&K (1986). This principle directly applies where sensor or AI failure in autonomous vehicles infringes the right to life or safety. Therefore, the distinction between damnum and injuria defines the boundaries of liability in autonomous vehicle contexts.

In determining liability for autonomous vehicles, tort law requires more than mere damage; a culpable mental element or fault remains relevant, although the principle actus non facit reum nisi mens sit rea does not apply as strictly as in criminal law. Generally, liability is grounded in wrongful intention or culpable negligence, where negligence denotes a breach of the duty of care. In Blyth v Birmingham Waterworks Co. (1856), the reasonable man standard was established, while Donoghue v Stevenson (1932) introduced the neighbour principle and reasonable foreseeability as the basis of duty. Subsequently, Bolton v Stone (1951) and The Wagon Mound (1961) demonstrate that liability arises only where the harm is reasonably foreseeable and proximate, although novus actus interveniens may break the chain of causation. However, in the context of autonomous vehicles, fault-based liability alone is insufficient, as AI-driven systems lack human intention. As a result, strict or absolute liability, as recognized in Rylands v Fletcher (1868), becomes relevant, where liability depends not on fault but on risk creation and control failure. In this context, liability becomes increasingly multi-layered, extending to the manufacturer, software developer, and system designer, while the Product Liability doctrine attributes responsibility for software defects, sensor failures, or algorithmic malfunctions to the producer. At the same time, Vicarious Liability is being reinterpreted on the basis of corporate control and code ownership. Therefore, in self-driving vehicles, liability is determined by duty of care, breach, foreseeability, causation, and risk allocation, indicating a shift from human fault to technological risk management.

In the case of autonomous vehicles, insurance law operates as a compensation-first mechanism, as its primary objective is to provide prompt financial redress to victims after an accident. Under the doctrine of subrogation, the insurer first compensates the victim and subsequently pursues recovery from the responsible manufacturer, software developer, or other third party, ensuring that the victim is not required to await the resolution of liability disputes. Similarly, the doctrine of contribution distributes compensation among multiple liable parties, which is particularly significant in determining shared liability in the context of autonomous vehicles. In addition, the principle of proximate cause identifies the actual legal cause of an accident, assisting in distinguishing liability among sensor failure, algorithmic error, or human override. Thus, in autonomous vehicle systems, insurance law shifts the focus from fault-based disputes to a compensation and recovery-based structure, providing a faster, fairer, and more effective solution in an era of technology-driven risk.

In the United States, the legal framework for autonomous vehicles remains a decentralized patchwork system, where the National Highway Traffic Safety Administration (NHTSA) provides safety guidance at the federal level. However, vehicle licensing, registration, traffic enforcement, insurance, and liability are primarily governed by individual states. Although the proposed SELF DRIVE Act and AV START Act seek to introduce limited federal preemption in areas such as design, construction, and performance regulation, a comprehensive national law has not yet been enacted. As a result, testing and deployment depend on state-based permits and certification. Nevada follows a self-certification model, while California permits public-road testing with a safety driver and, in certain cases, allows driverless operation. Consequently, liability for autonomous vehicles continues to be determined within traditional tort and insurance frameworks, with state-level control proving more effective in practice than a uniform federal rule.

The European Union adopts an integrated risk-based regulatory model for autonomous vehicles, built upon harmonized safety standards, data protection, and AI governance. Regulation (EU) 2019/2144 ensures accident traceability through advanced safety systems and mandatory event data recording. At the same time, the General Data Protection Regulation (GDPR) protects personal data collected by automated vehicles as a fundamental right and imposes strict controls on its processing. In addition, the EU AI Act (2024/1689) classifies autonomous vehicles as high-risk AI systems, requiring transparency, accountability, and human oversight. Thus, the EU framework prioritizes human rights, safety, and ethical regulation over purely technological advancement.

In the United Kingdom, the legal framework establishes a clear compensation-first liability model. Under Section 2 of the Automated and Electric Vehicles Act 2018, primary liability for accidents occurring in autonomous mode is imposed directly on the insurer. This allows victims to receive prompt compensation, while fault determination is deferred to a later stage. The insurer may subsequently recover losses from the manufacturer or other responsible parties through subrogation. This model reduces the complexity of liability adjudication and creates a victim-centric, efficiency-oriented system.

Germany’s legal framework brings autonomous vehicles under regulated autonomy, maintaining a balance between technological capability and legal responsibility. Amendments to the Strassenverkehrsgesetz (StVG) recognize automated driving and mandate data recording under Section 63a. This enables accurate determination of control at the time of an accident. The resulting traceability mechanism establishes a mixed liability regime, where responsibility may be attributed to both the driver and the manufacturer.

Japan’s legal framework for autonomous vehicles is based on staged authorization and strict safety compliance. Under the Road Traffic Act and the Road Transport Vehicle Act (2019 Amendment), Level 3 automation has been legally recognized, and Level 4 operation is permitted under specific conditions pursuant to Section 41. In this system, in cases of system failure, the driver remains liable as a fallback, ensuring continued human oversight within technological autonomy. Thus, Japan’s model reflects controlled innovation with a clear prioritization of safety. Japan aims to introduce autonomous shuttle services in major cities by 2030.

In India, the legal framework for autonomous vehicles remains within a regulatory vacuum, as the existing Motor Vehicles Act, 1988 is entirely driver-centric. Although Section 2B provides scope for the promotion of innovation, no specific liability or operational framework for autonomous vehicles has yet been established. As a result, India remains in a policy development phase, where technological adoption and legal regulation are evolving in parallel.

In the context of Bangladesh, the relevance of this technology has created a significant regulatory and liability gap, as existing driver-centric laws are inadequate to address AI-driven systems or software failures. Incidents such as recent cyber-related risks, including the potential misuse of AI-enabled or Tesla-like systems to weaponize vehicles, demonstrate that autonomous vehicles are no longer merely a mode of transport but complex cyber-physical systems. Therefore, to develop a sustainable legal framework, it is no longer optional but necessary to move beyond traditional negligence-based approaches and incorporate principles such as absolute liability, algorithm-based responsibility, and strict product security standards.

In the Miami Tesla Verdict (2025), a federal jury held Tesla partially liable for misleading presentation and design defects in its Autopilot system, thereby extending manufacturer liability beyond the scope of driver negligence.

In Huang v. Tesla (2024 – Settlement), the case was settled over allegations of sensor failure in the Autopilot system, highlighting the significance of software defects and design accountability in the autonomous vehicle context. In Arizona v. Uber (2018/2023), negligence of the safety driver in a self-driving vehicle accident was held responsible, underscoring the continued relevance of contributory negligence and human oversight. In re: Tesla ADAS Litigation (2025 – Ongoing) involves allegations of misrepresentation and breach of warranty regarding FSD capability, linking autonomous vehicle liability with the consumer protection framework.

From an analytical standpoint, liability in autonomous technology is no longer merely a question of outcomes; it is fundamentally shaped by system design, algorithmic logic, and code-based probability. Here, AI does not simply replace the driver; it displaces human intention itself, rendering the traditional reasonable man standard increasingly inadequate. From this perspective, AI presents a legal paradox: a non-human system is endowed with decision-making capacity, yet it is not recognized as a full legal entity. This gap is likely to become the central foundation for the future reconstruction of liability frameworks.

A prudent recommendation is the adoption of a calibrated utilitarian- deontological hybrid framework, integrating data traceability, cybersecurity safeguards, control attribution, AI-specific standards of care tailored to algorithmic decision-making systems, and international standardization within the regulatory architecture governing autonomous vehicles.

It is proposed that each autonomous vehicle be mandatorily equipped with an Event Data Recorder (black box) to determine whether human or machine control was operative at the time of an accident, thereby ensuring evidence-based liability. In autonomous systems, effective control holds greater legal significance than ownership. In adverse conditions, such as heavy rainfall or sensor unreliability, predefined operational protocols should require the system to enter a safe mode, including controlled deceleration or complete stop.

It is further suggested that multiple sensor fusion, compliance with safety standards such as ISO 26262, and the integration of ethical parameters within rigorous quality assurance processes be ensured, as extended accident-free operation cannot guarantee absolute safety.

It is also advisable that a heightened statutory duty of care be imposed on manufacturers to address cybersecurity risks, given that autonomous vehicles function as cyber-physical systems where system compromise directly affects road safety. In this context, alongside the reasonable man standard, a reasonable autonomous system standard should be established as a benchmark for safe system design.

It is recommended that policy formulation prioritize stakeholder consent and the broader public interest, with particular attention to those at risk of employment displacement through early rehabilitation and skill transformation measures.

Finally, it is advocated that, beyond fragmented national approaches, a harmonized international framework such as UN Regulation No. 157 be adopted to ensure consistent and sustainable governance of autonomous vehicle technology.

Bibliography

Books

Salmond JW, Jurisprudence (12th edn, Sweet & Maxwell).

Basu DD, Law of Torts (LexisNexis 2018).

Journal Articles & Academic Sources

Holstein T, Dodig-Crnkovic G and Pelliccione P, ‘Ethical and Social Aspects of Self-Driving Cars’ (2018) arXiv.

Bonnefon JF, Shariff A and Rahwan I, ‘The Social Dilemma of Autonomous Vehicles’ (2016) Science.

Lin P, ‘Ethics of Autonomous Cars’ (2016).

Web Sources

Waymo, ‘Self-Driving Technology Overview’ https://waymo.com accessed 2026.

Case Law

Acton v Blundell (1843) 12 M&W 324.

Dickson v Reuter’s Telegraph Co (1877) 2 CPD 62.

Day v Browning (1878) 10 Ch D 294.

Ashby v White (1703) 2 Ld Raym 938.

Bhim Singh v State of J&K AIR 1986 SC 494.

Blyth v Birmingham Waterworks Co (1856) 11 Ex 781.

Donoghue v Stevenson [1932] AC 562 (HL).

Bolton v Stone [1951] AC 850 (HL).

Overseas Tankship (UK) Ltd v Morts Dock (The Wagon Mound No 1) [1961] AC 388.

Rylands v Fletcher (1868) LR 3 HL 330.

Reports

National Highway Traffic Safety Administration (NHTSA), Automated Driving Systems: A Vision for Safety (US Department of Transportation).

Morgan Stanley, Autonomous Vehicles: Economic and Market Impact Report.

Legislation & Regulations

Intermodal Surface Transportation Efficiency Act 1991 (USA).

Automated and Electric Vehicles Act 2018 (UK).

Regulation (EU) 2019/2144 (General Safety Regulation).

Regulation (EU) 2024/1689 (Artificial Intelligence Act).

General Data Protection Regulation (EU) 2016/679.

Strassenverkehrsgesetz (StVG) (Germany).

Road Traffic Act and Road Transport Vehicle Act (Japan, 2019 Amendment).

Motor Vehicles Act 1988 (India).

UN Regulation No 157 (Automated Lane Keeping Systems).