Bridging the gap: A practical guide to generative AI in legal disclosure

As UK legal practitioners navigate the rise of generative artificial intelligence (GenAI), they are increasingly faced with a fundamental problem: how can it be used responsibly in court-ordered disclosure exercises when there is no clear legal or procedural playbook to follow?

With the International Legal Technology Association’s (ILTA) support, a working group of litigation and disclosure practitioners, comprising Fiona Campbell (Fieldfisher), Tom Whittaker (Burges Salmon), David Morgan-Wilkins (Norton Rose Fulbright), James MacGregor (ILTA; Ethical eDiscovery), and Jamie Tomlinson, Imogen Jones and Jonathan Howell (DAC Beachcroft), united to address this challenge. This union resulted in the shared drafting of the GenAI Practitioner Best Practice Guide (the GenAI guide), which was released by ILTA on 30 September 2025.

The GenAI Guide was initially prepared in late 2024 as an addendum to ILTA’s well-received Active Learning Best Practice Guide, published earlier that same year. The GenAI guide was drafted with adaptability in mind: while it references Practice Direction 57AD (PD57AD), it is intentionally broad, so that its recommendations may be applied in full or on an ’à la carte’ basis to suit different GenAI uses in legal practice.

Throughout 2025, amid successive rewrites of the draft GenAI guide, the working group observed GenAI use in disclosure evolve from a tool whose validation depended on deployment alongside Active Learning to a technology used independently in court-ordered disclosure under PD57AD, specifically through practitioner cooperation; this development occurred principally where opposing parties used comparable eDisclosure platforms with similar black-box foundational GenAI models. Accordingly, the final GenAI guide was published as a standalone best-practice document rather than an addendum. In finalising the document, the working group took into account comments from a two-month public consultation and combined those submissions with its own practical experience of GenAI in legal workflows and wider input from clients, technologists and the profession.

James MacGregor states: “PD57AD was developed following extensive collaboration between the lawyers and technologists who formed the Disclosure Working Group (DWG), which lead to the Disclosure Rules Pilot, before this was written into the Practice Direction in late 2022. In the few years since then, the adoption and advancement of AI has been exponential, which has demanded that a new consortium of lawyers and technologists expand on the work of the DWG to ensure the Practice Direction remains relevant in the post-ChatGPT era. The work being done by this group is to enable the English courts to empower those who choose to litigate in this jurisdiction with the ability to select advanced eDiscovery technology, when following the appropriate guidance.”

Why is there a need for the GenAI guide?

PD57AD, permits the use of Technology Assisted Review (TAR), placing predictive coding acceptance on a statutory footing in 2019 (originally through the Disclosure Pilot Scheme (PD51U)), following initial judicial approval by the High Court in the 2016 matter of Pyrrho Investments Ltd v MWB Property Ltd.

Considering that PD57AD was drafted before the advent of GenAI, the Practice Direction does not even begin to address how emerging technologies such as GenAI can or should be deployed in court-ordered disclosure. That absence of guidance has left many litigation teams hesitant to adopt GenAI independently, particularly given PD57AD’s emphasis on transparency: disclosure workflows must be consistent, auditable and defensible, and therefore may be open to scrutiny or challenge where disagreement arises.

To address this gap, the working group’s shared goal was to provide fellow practitioners with a practical resource that reflects both the capabilities and the risks of GenAI in court-ordered disclosure, moving litigation teams from theory to action.

Fiona Campbell says: “The guide provides a sensible, structured framework for parties to agree the potential use of GenAI tools at the outset, avoiding the need to revisit or amend the DRD further down the line. Even if GenAI is not ultimately used, agreeing the parameters in advance gives both sides clarity and comfort. It sets out a balanced approach that supports transparency, accountability, and defensibility, which are key ingredients for responsible innovation in disclosure.”

Principles and use cases

The GenAI guide outlines core GenAI use cases in disclosure, from redaction assistance and privilege flagging to issue classification and relevance prediction, while avoiding blind adoption. Its non-prescriptive approach promotes explainability, validation and human oversight, making clear that GenAI should augment rather than replace legal judgment. That alignment is especially important in light of the Divisional Court’s decisions in Ayinde v London Borough of Haringey and Al-Haroun (June 2025), which confirmed that solicitors and barristers remain fully accountable for AI-generated content and may face regulatory sanction if they fail to verify its accuracy. Recognising variation in how firms and clients are approaching GenAI, the working group therefore offered a flexible set of best practices rather than rigid rules. The guide includes suggested workflows for prompt testing, continuity of evidence governance and methods for integrating GenAI into existing Active Learning review strategies, so practitioners can apply its recommendations proportionately to their needs.

Tom Whittaker says: “This guide aims to help practitioners turn principles into practice, achieving responsible innovation that is compliant with PD57AD.”

David Morgan-Wilkins says: “Disclosure as an area of practice is well-suited to benefit from recent advancements in the field of AI and machine learning. However, concerns about the defensibility of such technologies and the absence of standard consensus as to their use has meant they are often avoided in favour of more familiar, manual approaches. Following on from the release of ILTA’s Active Learning guide in 2024, which sought to address these concerns for ‘TAR 2.0’ analytics, this document provides a foundational framework for the use of GenAI in document review. It is hoped that the existence of such a framework at this early stage in the development of this technology will encourage its uptake while ensuring appropriate safeguards are adopted. This is an opportunity to make disclosure cheaper, more effective and more accessible.”

A collaborative effort

The development of the guide itself reflected the principles the working group wanted to promote: cross-firm cooperation, transparency, and collective knowledge-building. From the outset, the working group ensured that the drafting process was neutral, vendor-agnostic, and inclusive of diverse views across legal and technical disciplines.

Jamie Tomlinson says: “This guide is a testament to what can be achieved by lawyers from multiple firms pulling in the same direction. The regulatory lacuna left by the explosion of GenAI requires a practical, future-proof guide to help litigators of all specialisms navigate the new technological landscape. That could only have been achieved with the range of skills and knowledge that this team brought to the table. Each contributor brought a unique perspective: technical insights into how GenAI tools behave in review environments, detailed knowledge of procedure and regulation, and first-hand experience of the operational challenges faced by busy litigation teams.”

Imogen Jones states: “The guide emphasises the need for structured testing, validation and proper oversight in the application of GenAI, which is key to ensuring the integrity of the current disclosure process is maintained and to ensure that there are measurable outputs which the court can consider. GenAI relies on non-deterministic algorithms and therefore proper testing and validation allows parties to demonstrate defensibility.”

The working group also considered the wider strategic question of how litigation teams can scale GenAI safely while maintaining consistency across matters and ensuring workflows would withstand judicial scrutiny.

Jonathan Howell says: “As legal practitioners navigate the evolving e-disclosure landscape, the key is striking a balance between embracing GenAI innovations and maintaining defensibility. From the outset of the project the contributors have recognised that the future scalability of GenAI within legal teams depends not just on the underlying technology itself, but on building workflows that are transparent, auditable and adaptable to regulatory shifts.”

What comes next?

What comes next is practical and iterative. Practitioners using GenAI should consider incorporating the GenAI guide’s principles into case management at an early stage, thereby agreeing scope and limits between parties so that the chosen approach for GenAI use is accepted rather than disputed. The working group will monitor judicial and regulatory developments and publish revisions via ILTA as necessary; the Guide is intended as a living framework, open to adaptation as courts, regulators and professional bodies clarify expectations and legislative guardrails for GenAI.