In large pieces of group litigation, there is often a significant amount of documentary evidence to collate and share in a disclosure process. This can be a lengthy and costly process for clients. At DWF, we are embracing the use of Generative AI to accelerate disclosure exercises and reduce cost.
The governing procedural framework is, of course, primarily the Civil Procedure Rules (“CPR”), in particular CPR 31 and, in the Business and Property courts, including the Technology and Construction court, PD 57AD. The CPR did not anticipate the use of Generative AI, but PD 57AD does provide for the use of technology-assisted review software in the disclosure process.
Judges, regulators and practitioners will clearly have to grapple with the implications of the use of AI. There is limited legal authority to assist. The Civil Justice Council1 is currently examining the use of AI for preparing court documents; one aspect of this examination attaches to disclosure. In an interim report, published in June 2026, the CJC noted that they are currently consulting on whether there is a need to introduce a section within the disclosure list/statement addressing the extent to which AI tools have been used. We will report on this part of the consultation, and indeed the wider consultation on the use of AI for preparing court documents later in 2026 when the final report is published.
The International Legal Technology Association (“ILTA”) has published a set of principles to guide practitioners. This article distils those principles—to outline how legal teams can responsibly deploy Generative AI (GenAI) within disclosure workflows.
Transparency: Clear Disclosure of Methods and Use of GenAI
Across all guidance, transparency is the single most consistently emphasised requirement.
- The ILTA Generative AI Best Practice Guide stresses early and open discussions between parties about why and how GenAI tools will be used, recommending that reasoning and workflows be captured in procedural documents such as, in the Business and Property Court, the Disclosure Review Document (DRD), governed by PD57AD.
- Personal Injury practitioners should consider how best to seek such agreement with opponents, probably by way of correspondence prior to the date for disclosure, to agree parameters in advance.
- UK judicial commentary increasingly frames GenAI governance, more widely, as a matter of professional responsibility, as illustrated in Ayinde v London Borough of Haringey, where accountability for inaccurate AI‑generated material was placed squarely on the human legal team.
Practical takeaway: Liaise with opponents, document the workflow, disclose methodologies where appropriate, and ensure both opponents and the court can understand the role GenAI plays in document review.
Defensibility: Validation, Testing and Audit Trails
Parties using GenAI are responsible for ensuring that their workflows are robust, reliable, and reproducible.
- ILTA guidance emphasises quality assurance techniques, including precision/recall testing, random sampling, iterative validation, and maintaining full audit logs of prompts and outputs.
- The Law Society echoes the need for methodology that withstands judicial scrutiny, noting that GenAI adoption must “follow appropriate guidance” to remain defensible under CPR.
Practical takeaway: Build defensibility in from the outset—plan for transparency, testing, and documentation before deploying GenAI on substantive review tasks. Key stakeholders should have a clear understanding of the process, the risks and the points at which decisions are made.
Human Oversight: AI as a Tool, Not a Substitute
All sources reaffirm that GenAI must support—not replace.
- ILTA guidance emphasises accountability and the human‑led nature of GenAI‑enabled workflows.
- In practice, this means that lawyers must define issues, set parameters, and supervise methodology; GenAI cannot autonomously satisfy these requirements.
Practical takeaway: Ensure qualified lawyers, and in particular the subject matter experts, remain directly involved in a) scoping issues, b) defining, testing, and tuning prompts and c) conducting blind validations.
Privacy and Confidentiality Controls
There may be some nervousness among clients relating to the use of GenAI and the potential exposure of sensitive or confidential client information.
- The Law Society has indeed highlighted that CPR, in pre‑dating GenAI, does not address confidentiality issues arising from use of emerging AI tools, making additional vigilance essential.
- Professional responsibility guidance warns against inputting confidential data into systems without proper safeguards or clarity around data usage.
In a recent decision of the Upper Tribunal (Immigration and Asylum Chamber)[1], the tribunal indicated that uploading confidential material into publicly available AI platforms may be treated as placing that information into the public domain, with the result that client confidentiality is lost and any associated claim to legal professional privilege may fail.
Practical takeaway: Conduct a security and privacy assessment of GenAI tools before use, to understand where data is sent for analysis and ensure that controls are in place to prevent the underlying model retaining, learning from, or re‑using confidential inputs.
Identifying Unsuitable Documents and Use Cases
Not all disclosure tasks are appropriate for GenAI treatment.
- The ILTA Guide recommends early identification of documents or issues not suitable for GenAI, preventing inefficiencies or risks.
- Examples include nuanced determinations in relation to privilege, and complex documents requiring specialist human interpretation, such as large financial spreadsheets or highly technical material.
- GenAI analysis is often based on text; risk is introduced if the text does not accurately reflect the original document, such as poorly scanned documents or documents that contain tables, diagrams or handwritten annotations.
Practical takeaway: Perform analysis and classify datasets early to determine what can safely be handled by GenAI and what requires traditional technology assisted review or manual review.
Evolving Regulatory and Judicial Landscape
GenAI use is occurring against a backdrop of quickly shifting regulatory expectations:
- The Law Society notes that PD57AD is currently under pressure to evolve, with GenAI now widely used during disclosure despite no explicit procedural rulebook.
- UK judges are increasingly signalling that appropriate AI usage must fit within established duties of competence, verification and candour.
Practical takeaway: Remain alert to forthcoming judicial statements and updates to PD57AD and CPR 31 or related practice notes—2026 is expected to bring more explicit guidance.
Conclusion
GenAI presents opportunities for accelerated, more cost‑effective disclosure, but only when deployed in alignment with fundamental principles: transparency, defensibility, human oversight, confidentiality controls, and task suitability.
GenAI should and can sit comfortably within disclosure workflows, but only where it enhances—not erodes—the disciplined, cooperative and defensible processes that any disclosure regime demands.
[1] UK v Secretary of State for the Home Department [2026] UKUT 81 (IAC)
