The Pitfalls of Using AI in UK Civil Litigation

AI in civil litigation: promise and peril

Artificial intelligence is increasingly being integrated into legal processes, promising efficiency, lower cost and improved access to justice. In the civil courts of England and Wales, AI-related tools are already used for tasks such as document review, disclosure and legal analysis.

Technology-assisted review and predictive coding provide the clearest established example. In Pyrrho Investments Ltd v MWB Property Ltd [2016] EWHC 256 (Ch), the High Court approved predictive coding in e-disclosure as a proportionate and efficient approach. Since then, AI-supported review tools have become part of the wider move towards digital litigation.

The central challenge is not whether AI should be used at all. It is whether AI can be used without weakening the foundations of civil justice: open reasoning, equality of arms, judicial independence, procedural fairness and the right to be heard.

Bias in AI decision-making

Algorithmic bias is one of the most serious risks in legal AI. AI systems learn from data, prompts, design choices and human instructions. Where those inputs reflect existing inequality, the output may reproduce or amplify it.

In civil litigation, that risk could arise in case prediction, settlement modelling, triage, disclosure prioritisation, damages assessment or credibility analysis. A system trained on historical outcomes may absorb historic disadvantage and present it as neutral probability.

The Court of Appeal’s decision in R (Bridges) v Chief Constable of South Wales Police [2020] EWCA Civ 1058 is instructive. Although the case concerned automated facial recognition rather than civil litigation, it underlined the need for public authorities to assess equality impacts and potential technological bias. The Public Sector Equality Duty under section 149 of the Equality Act 2010 remains relevant where public decision-makers deploy new technology.

UK GDPR also reinforces this point. Article 22 addresses automated decision-making with legal or similarly significant effects, while Recital 71 warns against discriminatory effects and requires appropriate safeguards to minimise errors.

Confirmation bias and automation bias

AI does not only carry its own risks. It can intensify human cognitive bias. Confirmation bias is the tendency to favour information that supports an existing view. Automation bias is the tendency to over-trust machine output because it appears objective, confident or precise.

In litigation, those risks matter. A lawyer may accept an AI-generated case summary because it supports their client’s position. A litigant in person may rely on AI-generated authorities without checking them. A judge may be subtly influenced by a machine-generated summary, even where the final decision remains formally human.

The judicial guidance issued in December 2023 warns judicial office holders that AI-generated information may be inaccurate, out of date, fictitious or biased, and must be independently verified. It treats generative AI as a secondary tool, not a replacement for legal research, professional judgment or judicial reasoning.

The tax tribunal case referred to in the draft, Harber v HMRC, illustrates the risk. A litigant in person relied on AI-generated case citations that were not real. The problem was not malicious fabrication. It was misplaced trust in an automated answer.

Threats to due process and procedural fairness

The deepest concern is due process. Civil justice depends on impartial adjudication, transparent reasoning, the ability to know and answer the case, and decisions made by lawful authority on proper evidence.

AI can disturb those principles where it is opaque, unchallengeable or embedded into decision-making without disclosure. A party cannot meaningfully contest a process they cannot see or understand.

The judicial guidance permits cautious use of AI for appropriate administrative or drafting support, but not as a substitute for substantive legal analysis. That distinction is essential. Efficiency cannot justify a process in which affected parties cannot understand the basis of a decision.

Regulatory framework: GDPR, the EU AI Act and judicial guidance

The UK regulatory framework is developing through several overlapping routes: data protection law, equality law, professional duties, judicial guidance and broader AI policy. The resulting picture is useful but fragmented.

The judicial AI guidance for England and Wales is particularly important because it speaks directly to the courts. It warns judges about hallucinations, bias, confidentiality, security and the limits of generative AI. It also confirms that lawyers and parties remain responsible for material filed with the court.

Professional duties also matter. Solicitors and barristers using AI remain subject to duties of competence, candour, confidentiality and responsibility to the court. A lawyer cannot avoid responsibility by saying an AI tool produced the error.

Implications for litigants in person

Litigants in person sit at the sharp edge of legal AI. AI may help them understand law, draft documents and navigate procedure. It may also mislead them with plausible but false answers, invented authorities or overconfident predictions.

The legal system should not assume that AI automatically improves access to justice. It may do so only if tools are accurate, explainable, affordable, accessible and paired with human support. For LiPs, the danger is a new procedural trap: a tool that appears to help but quietly creates defective pleadings, false authorities or misplaced confidence.

Practical safeguards could include vetted public legal AI tools, plain-language warnings, court guidance for LiPs, judicial tolerance for genuine AI-related mistakes, and clear rules requiring verification of all authorities and quotations.

Conclusion: innovation must remain answerable to fairness

AI in civil litigation is a double-edged development. It can reduce cost, speed up document review and support access to justice. It can also import bias, amplify human overreliance, generate false law and weaken the transparency on which fair process depends.

The UK framework is moving in the right direction, but remains patchwork. UK GDPR, equality law, the EU AI Act, judicial guidance, professional duties and policy principles all point towards the same requirements: human oversight, transparency, accountability, non-discrimination and contestability.

For litigants in person, the stakes are especially high. AI can democratise access to legal assistance, but only if the system guards against false confidence, unequal access and opaque decision-making. The future of AI in the courts should therefore be measured not by technical novelty, but by whether it preserves the right to a fair, intelligible and humanly accountable process.