AI‑Autonomous Vehicles: Civil Liability and Regulatory Frameworks

Exploring how courts, statutes, and emerging policies allocate responsibility when driverless cars cause harm.

1. Foundations of Civil Liability for Motor‑Vehicle Harm

Modern tort law still rests on the principle that a negligent driver who causes injury is liable for damages. The classic framework was articulated long before autonomous systems existed. In Miller v. Civil City of South Bend (7th Cir. 1990‑05‑24, Docket 88‑3006), the Seventh Circuit reaffirmed that a municipality may be held liable for failing to maintain safe road conditions, applying the ordinary negligence standard to public‑entity duties. Although the case predates autonomous‑vehicle (AV) technology, its analysis of duty, breach, causation, and damages provides the baseline against which newer regimes are measured ([Miller v. Civil City of South Bend]).

When an AV is involved, the “driver” is a software‑controlled system, raising the question of who satisfies each tort element. The literature consistently notes three potential fault sources: (1) the vehicle owner/operator, (2) the vehicle manufacturer or software developer, and (3) third‑party infrastructure providers. Each source may trigger a different liability theory—negligence, strict product liability, or premises liability—depending on the jurisdiction’s statutory and case law. Understanding how traditional principles are being reshaped is essential for anyone building, deploying, or regulating autonomous fleets.

2. Korean Civil Liability Landscape

South Korea has become a focal point for scholarly analysis of AV liability. Kim Hyun‑soo’s 2017 article “Autonomous Vehicle Accident and Civil Liability” examines how Korean civil law, which traditionally requires a human driver’s fault, must adapt to AI decision‑making. Kim argues that the existing negligence framework is strained because the “driver” is an algorithm, and courts will need to assess the reasonableness of the vehicle’s programmed behavior rather than a human’s conduct ([Kim Hyun‑soo 2017]).

Building on that foundation, Bong Geun Shin (2018) expands the discussion to product‑liability concepts, suggesting that manufacturers could be held strictly liable for design defects in autonomous systems, even absent proof of negligence. Shin notes that Korean statutes already impose strict liability for defective products, and that AVs may fall squarely within that regime ([Shin 2018]).

More recently, Park and Ahn (2024) provide a comprehensive “Civil Liability Regime on Autonomous‑Driving Vehicle Accident in Korea.” Their analysis maps out three liability channels: (1) owner/operator liability when the vehicle is manually overridden, (2) manufacturer liability for software errors, and (3) infrastructure liability where road‑sign misreadings cause accidents. Park & Ahn recommend legislative clarification to define the “reasonable AI behavior” standard, arguing that without statutory guidance, courts will face inconsistent applications of negligence and product‑liability doctrines ([Park & Ahn 2024]).

Collectively, these Korean studies highlight a transitional legal environment: courts are beginning to apply existing tort concepts, but scholars call for explicit statutes that articulate the duties of AI developers and the expectations placed on autonomous systems.

3. German Approach to Autonomous‑Vehicle Liability

Germany’s civil code (BGB) already contains robust product‑liability provisions, and scholars have examined how these translate to AVs. Martin Ebers (2025) surveys the German legal landscape in “Civil Liability for Autonomous Vehicles in Germany.” Ebers observes that German courts are likely to treat an autonomous car as a “product” under §§ 823 and § 1 of the Product Liability Act, thereby imposing strict liability on manufacturers for defects that cause injury ([Ebers 2025]).

Ebers also points out that German law retains a “fault‑based” component for owners who fail to maintain the vehicle or ignore software updates. In such cases, the owner may be liable under ordinary negligence principles, similar to the traditional driver‑fault model. The article stresses that Germany’s dual‑track system—strict liability for manufacturers plus negligence for owners—offers a clearer allocation of responsibility than jurisdictions still wrestling with how to fit AVs into existing frameworks.

4. Comparative Tort Frameworks Across Jurisdictions

The 2026 World Electric Vehicle Journal article “Comparing Tort Liability Frameworks in Autonomous Vehicle Accident Governance” by Bo Long, Ziyu Zhao, and Qianyi Cai provides a cross‑jurisdictional survey of how different legal systems are handling AV torts. The authors categorize approaches into three archetypes:

Strict‑Product‑Liability Dominant – exemplified by Germany and several EU states, where manufacturers bear primary responsibility for design flaws.
Negligence‑Shift to Operators – seen in parts of East Asia (including Korea), where owners may be liable if they fail to supervise or update the vehicle.
Hybrid or “Shared‑Risk” Models – emerging in the United States, where courts blend negligence, product liability, and emerging statutory schemes (e.g., state‑level autonomous‑vehicle statutes).

Long et al. argue that the hybrid model aims to balance innovation incentives with victim compensation, but it also creates uncertainty for manufacturers operating across state lines. Their comparative chart (not reproduced here) shows that no single framework has achieved consensus, underscoring the need for harmonized international guidelines.

5. Criminal Liability When an AV Causes Harm

Beyond civil remedies, autonomous systems raise novel criminal‑law questions. The Santa Clara Law Review article “Criminal Liability Issues Created by Autonomous Vehicles” (date unspecified) explores scenarios where an AV’s autonomous decision‑making could satisfy the mens rea element of a crime. The authors caution that traditional doctrines—recklessness or negligent homicide—may be stretched to accommodate software‑driven conduct, especially when a manufacturer’s known defect leads to a fatal crash.

Kristina Sashikovna Bogdanova’s 2024 monograph “Issues of Criminal Liability When Causing Harm by an Autonomous Vehicle” delves deeper, proposing three criminal‑liability pathways: (1) Corporate criminal liability for knowingly releasing unsafe software, (2) Operator liability when manual overrides are misused, and (3) Strict liability offenses for certain traffic violations where fault is irrelevant (e.g., driving under the influence of a malfunctioning AI). Bogdanova emphasizes that most jurisdictions lack statutory language to prosecute AI‑driven conduct, leaving prosecutors to rely on analogies to existing offenses ([Bogdanova 2024]).

Both works agree that without clear legislative guidance, criminal prosecutions will be inconsistent, potentially chilling AV development or leaving victims without adequate redress.

6. Emerging Regulatory Responses and Practical Guidance

Regulators worldwide are beginning to fill the legislative vacuum. While the Federal Register notice on the EPA’s rescission of greenhouse‑gas standards (2024) does not directly address AVs, it illustrates how agencies can overturn longstanding rules, reminding AV stakeholders that regulatory stability is not guaranteed ([EPA 2024]). In the AV context, several jurisdictions have introduced autonomous‑vehicle testing permits that require insurers to carry minimum coverage, and some states (e.g., California, Michigan) have enacted statutes defining “operator” responsibilities for software updates.

Practical guidance distilled from the scholarly record includes:

Document Software Versions – Park & Ahn (2024) stress that manufacturers should keep immutable logs of firmware versions deployed in each vehicle, facilitating fault tracing in litigation.
Implement Over‑the‑Air (OTA) Update Policies – Kim (2017) notes that failure to provide timely safety updates could be deemed negligent under Korean law.
Maintain Clear Owner‑Operator Agreements – Shin (2018) recommends contracts that allocate responsibility for manual overrides and routine maintenance, reducing ambiguity in negligence claims.
Conduct Risk‑Based Testing – Ebers (2025) highlights Germany’s requirement for “functional safety” certification (ISO 26262) before market entry, a practice that can mitigate product‑liability exposure.
Engage with Regulators Early – Long et al. (2026) advise firms to participate in state‑level rulemaking to shape hybrid liability frameworks that balance innovation with consumer protection.

7. Checklist for AV Stakeholders

| ✅ Action | Why It Matters | Primary Source | |---|---|---| | Maintain immutable firmware logs | Enables traceability for product‑liability claims | Park & Ahn 2024 | | Provide OTA safety updates promptly | Avoids negligence findings (Korean context) | Kim 2017 | | Draft explicit owner‑operator agreements | Clarifies negligence vs. product‑liability allocation | Shin 2018 | | Obtain functional‑safety certification (ISO 26262) | Meets German strict‑liability standards | Ebers 2025 | | Secure minimum liability insurance per jurisdiction | Satisfies emerging state statutes and protects against tort damages | Long et al. 2026 | | Participate in local regulatory consultations | Influences hybrid liability frameworks | Long et al. 2026 | | Develop a criminal‑liability response plan | Addresses potential corporate or operator prosecutions | Bogdanova 2024 | | Monitor regulatory changes (e.g., EPA, state AV laws) | Anticipates shifts that could affect compliance costs | EPA 2024 |

8. Maintaining Ongoing Compliance

Liability regimes for autonomous vehicles are still evolving. To stay compliant:

Quarterly Legal Audits – Review recent case law (e.g., new