--- title: "Article: Automated vehicles, artificial intelligence and the courts" canonical: "/article-automated-vehicles-artificial-intelligence-and-the-courts" published: 2026-11-13 --- # Article: Automated vehicles, artificial intelligence and the courts ## Automated vehicles, Artificial Intelligence and the courts #### Introduction AI is central to AV systems. It enables them to perceive their surroundings, process information and make driving decisions in real time. The AI used in these processes generates vast quantities of data, which is likely to become important in litigation. Although the courts have a good track record of adapting to technological change, AI presents a number of unique challenges, especially in claims arising from incidents involving AVs. “The data may establish what happened, but not necessarily why it happened.” #### The role of AI in AVs AI is responsible for how AVs drive and interact with other road users. AVs continuously gather information about their environment by using sensors, cameras, radar, lidar and onboard computing systems. Algorithms analyse that information to identify hazards, predict risks and determine appropriate driving responses. Unlike human drivers, machines will not have 'accidents'. Instead, as Chris Moore of Apollo ibott stated in our podcast on coverage, AV actions will be the result of intentional decisions, many of which will be made by AI. AVs record extensive operational information, including speed, steering inputs, braking activity, sensor outputs and environmental conditions. These records form part of the decision-making process used by the vehicle. The data used by AVs and the logic AI systems used to apply that data in decision-making processes will often be vital for deciding liability. Modern machine-learning systems operate through highly complex models containing millions of parameters. Even their developers may struggle to explain precisely how a particular outcome was produced. Determining why an AV acted in a specific manner may therefore be considerably more complex than evaluating the actions of a human driver. “AI-generated evidence will need to be understandable, scrutinised and tested.” #### Litigation issues involving AI-generated data At present, courts resolving road traffic disputes focus on the conduct of human drivers, relying on witness evidence, CCTV footage and expert reports as the primary sources of evidence. In AV-related litigation, attention will inevitably shift towards datasets, software systems and system performance. This will present several challenges to the courts and their users. Data integrity is unsurprisingly a significant concern. Parties seeking to rely upon AV-generated data must be able to demonstrate that the information has been accurately preserved and remains unaltered. Questions may arise regarding software updates, cybersecurity incidents, data corruption, storage failures or deliberate manipulation. As with other forms of digital evidence, establishing a clear chain of custody is likely to be important. A major challenge concerns transparency. Many AI systems operate as 'black boxes', generating outputs without providing readily understandable explanations of how those outputs were produced. Where an AV brakes, accelerates or changes direction before an incident, the available data may establish what happened but not necessarily why it happened. This is likely to create difficulties for courts attempting to assess liability or causation. Judicial concerns regarding AI-generated evidence can be seen in Abbott and Others v Ministry of Defence \[2026\] EWHC 941 (KB). In expressing his concerns relating to a large language model used to diagnose noise-induced hearing loss, Garnham J stated that "where a method cannot be understood, scrutinised or tested, it is unsuitable as a basis for judicial fact-finding". Accepting a conclusion generated by an opaque model without understanding how it was reached would, in his view, amount to a 'complete dereliction of duty'. Abbott demonstrates that courts are unlikely to accept complex technological evidence at face value merely because it is generated by sophisticated systems. Instead, the focus remains on explainability, transparency and evidential reliability. The decision provides a useful analogy for future AV litigation, where concerns about algorithmic opacity and the 'black box' nature of decision-making may be particularly acute. As part of the need to analyse AV data, expert evidence is likely to become increasingly important. Judges will often require specialist assistance to interpret operational data, software logs and machine-learning outputs, increasing costs and complexity. Disclosure raises further challenges. Relevant evidence may be held not only by vehicle owners but also by manufacturers, software developers, fleet operators, insurers and cloud service providers. Potentially relevant material may include event data recorder logs, sensor records, software version histories, safety assessments, system testing documentation and remote server logs. The courts are already familiar with managing large volumes of electronic evidence. Parties must identify, preserve and disclose relevant electronic material through proportionate searches. In AV litigation, those obligations may extend to include extensive datasets, metadata, software logs and machine-generated records. The volume and technical nature of such information will exceed anything that has been encountered to date in most current personal injury and road traffic litigation. The costs associated with identifying, reviewing and producing these materials may also create proportionality concerns. Extensive disclosure exercises are often expensive and may be difficult to justify in lower-value road traffic claims. #### Practical issues in the lower courts While senior courts may be capable of managing technically complex evidence, significant practical difficulties are likely to arise in the lower courts, where the vast majority of disputes are resolved. County Court proceedings are intended to provide proportionate and cost-effective dispute resolution. AV litigation may require specialist expert evidence, technical analysis and substantial disclosure exercises. In lower-value claims, the cost of analysing vehicle-generated data alone may exceed the amount claimed. Judicial familiarity with the underlying technology presents a further challenge. Judges are accustomed to evaluating witness evidence, documentary records and traditional expert reports. Most will have limited experience of machine-learning systems, software architecture or algorithmic decision-making. Time and resource constraints may exacerbate these difficulties. Lower courts routinely manage substantial caseloads and operate under considerable time pressure. Detailed analysis of source code, audit logs, system design and regulatory compliance may be difficult to accommodate within standard case management and hearing timetables. Electronic disclosure is likely to present particular challenges. A single AV can generate vast quantities of data during routine operation. Determining relevance, identifying appropriate search criteria and resolving disputes concerning access to commercially sensitive information may add considerably to the burden placed on courts and litigants. “In lower-value claims, the cost of analysing vehicle-generated data alone may exceed the amount claimed.” #### Possible solutions Addressing these challenges will likely require a combination of regulatory, procedural and technological reforms. First, greater standardisation of data collection and retention would improve evidential reliability. The information-gathering framework established by the Automated Vehicles Act 2024 provides a basis for developing consistent standards relating to record-keeping, audit trails and safety-related information. Standardised data formats would make evidence easier for courts, insurers and experts to interpret. Secondly, emphasis should be placed on explainable AI systems capable of providing intelligible explanations for their decisions which would assist courts when determining liability and causation. Explainability may also address some of the concerns identified in Abbott, where judicial acceptance depended upon the ability to understand and scrutinise the underlying methodology. Thirdly, judicial education will be essential. Specialist training could improve understanding of AI systems, software design and digital evidence. Similar initiatives have already assisted courts in addressing issues relating to cybercrime and electronic disclosure. Dedicated procedural guidance would also be beneficial. New practice directions, or amendments to existing guidance, could specifically address AI-generated evidence, machine-learning systems, source code disclosure and data preservation obligations. Early identification of relevant datasets and electronic sources could improve efficiency and reduce disclosure disputes. Finally, courts may make greater use of single joint experts and court-appointed technical advisers. Such measures could reduce costs while providing judges with independent technical assistance in appropriate cases. “In lower-value claims, the cost of analysing vehicle-generated data alone may exceed the amount claimed.” “The volume and technical nature of AV evidence could exceed anything currently encountered in most road traffic litigation.” #### Conclusion AI is fundamental to the operation of AVs and will inevitably shape future litigation arising from their use. Although AI-generated data has the potential to provide valuable and objective evidence, significant concerns remain regarding transparency, admissibility, disclosure and reliability. These challenges are likely to be most acute in the lower courts, where resource constraints and proportionality considerations may limit the ability to engage with highly technical material. The legal system has repeatedly demonstrated its ability to adapt to technological change. Through improved standards, enhanced judicial training, clearer procedural guidance, proportionate disclosure mechanisms and greater emphasis on explainable AI, the courts can develop effective methods for resolving AV-related disputes. Achieving that balance will be essential to ensuring that technological innovation progresses alongside legal accountability and that those harmed by AV systems receive the compensation they deserve without suffering undue delays or stress. Michael McCabe Solicitor DAC Beachcroft, CSG **Next page >>** **Careers** **Complaints / Our Policies** **Accessibility** **Legal & Regulatory** **Emergencies** **Modern Slavery Act Statement** 2025 DAC Beachcroft LLP. All rights reserved.