This guide addresses the responsible use of generative AI in electronic disclosure, focusing on defensibility under PD 57AD.
Guide · 17 pages · 22 min read · Published 2026-08-30
* Starts with "Designed for UK legal practitioners..." * *Never quote filenames or headings?* Checked. No quotes or raw heading titles used. * *No em dashes and no en dashes?* Checked. No dashes used except hyphenated words ("first-pass", "e-discovery"). * *Return ONLY paragraph text?* Yes.Designed for UK legal practitioners navigating modern electronic disclosure, this text addresses the core operational challenge of deploying generative AI models that analyse large document sets at scale yet risk introducing hallucinations or invalid concessions. It establishes a clear framework for defensibility under Practice Direction 57AD while incorporating judicial directions, Law Society guidance, ILTA standards, and rulings such as Ayinde. Readers are guided through appropriate use cases, including first-pass relevance filtering, privilege identification, redaction support, and document summarisation, contrasting these methods with established Technology Assisted Review workflows. Methodological rigour is maintained through prompt engineering, drafting, testing, systematic versioning, and audit sampling against human judgement. Furthermore, the text covers data protection compliance, Disclosure Review Document paragraph drafting, protocol transparency, and underlying source file architecture. By identifying
Read this guide on your phone, browse guides by topic or go back to the full PDF library.
17 pages · 22 min read
This guide explores the responsible application of generative AI in electronic disclosure, covering prompt engineering, human verification, and defensibility.
Published by Computer Forensics Lab on 2026-08-30. Original material of the practice, free to read, cite and download. The authority behind this subject is Practice Direction 57AD, paragraph 14, privileged documents, which you should read alongside this guide. See every guide's author and source.
Prefer a PDF that matches this page exactly? Download the current text as a PDF, generated from the current wording of the guide, including any later corrections.
GENERATIVEAIINDISCLOSURE · A GUIDE FOR UK LAWYERS Using Generative AI Responsibly in Electronic Disclosure Prompted Review, Human Verification and Defensibility Under PD 57AD COMPUTER FORENSICS LAB
§ ABOUT THE AUTHOR PREPARED BY COMPUTER FORENSICS LAB E-DISCOVERY TEAM ESTABLISHED 2007 · LONDON ISO 17025-ALIGNED PROCEDURES GENAI-ASSISTED REVIEW WORKFLOW S
§ CONTENTS In this guide 01 Executive summary 02 The problem in plain English: a reader that never tires and some time s invents 03 The guidance landscape: ILTA, the Law Society, the judiciary and Ayinde 04 Where GenAI fits: relevance, privilege, summarisation and redaction support 05 Prompt engineering as methodology: drafting, testing and version in g 06 Verification: sampling AI outputs against human judgement 07 Transparency, data protection and the protocol 08 Source architecture: where else the evidence lives 09 Worked examples 10 Common mistakes and technical limitations 11 Questions to ask · Suggested wording 12 Checklist and red flags · When to involve a digital forensic expert 13 Frequently asked questions 14 Glossary · References · Disclaimer · How a specialist laboratory can assist
§ 01 · ORIENTATION Executive summary THE HEADLINE POINT: GENAIREADSANDCHARACTERISESATSCALEUNDER INSTRUCTIONSYOUWRITE: THEINSTRUCTIONSARETESTEDLIKESEARCHTERMS, THE OUTPUTSVERIFIEDLIKEAJUNIOR ' SWORK, ANDTHELAWYERREMAINSACCOUNTABLE FOR ALL OF IT
§ 02 · FIRST PRINCIPLES The problem in plain English: a reader that never tires and some time s invents
§ 03 · THERULESOFTHEROAD The guidance landscape: ILTA, the Law Society, the judiciary and Ayinde
§ 04 · THE USE CASES Where GenAI fits: relevance, privilege, summarisation and redaction support
§ 05 · INSTRUCTION S AS METHOD Prompt engineering as methodology: drafting, testing and version in g
§ 06 · CHECKINGTHEREADER Verification: sampling AI outputs against human judgement
§ 07 · THE PAPER WORK AND THE DATA Transparency, data protection and the protocol
§ 08 · THEWIDERMAP Source architecture: where else the evidence lives QUESTION DECISION QUALIT Y PROMPT VERIFICATION PROCESSING SCOPE FILE RECORDS INVENTORY LAYERS HUMAN TEXT- LOG LAYER
§ 09 · IN THE WILD Worked examples EXAMPLE1 · THEPROMPTEDFIRSTPASSTHATSURVIVEDITSAUDIT EXAMPLE2 · THESUMMARYTHATINVENTEDACONCESSION EXAMPLE3 · THEPRIVILEGEFLAGSTHATACCELERATED, ANDNEVERDECIDED
§ 10 · WHEREITGOESWRONG Common mistakes and technical limitations Common mistakes Technical limitations
§ 11 · INTERROGATORIES & DRAFTING AIDS Questions to ask · Suggested wording Ask your client Ask your opponent Ask your e Discovery / forensic provider SUGGESTED WORDING · GENAI PA R AG RAPHFORTHEDRD / PROTOCOL
§ 12 · QUICK CONTROL Checklist and red flags · When to involve a digital forensic expert The GenAI checklist Red flags When to involve a digital forensic expert
§ 13 · COMMON QUESTIONS Frequently asked questions Is it actually permissible to use generative AI in court-ordered disclosure? How is this different from the TAR we already use? What is "hallucination" and how do we protect against it in disclosure work? Do we have to tell the other side we used AI? Can AI make privilege calls? Our client's data is highly sensitive. Does that rule GenAI out?
§ 14 · REFERENCE Glossary Sources and authoritative references DISCLAIMER
§ HOW A SPECIALIST LABORATORY CAN ASSIST Working with Computer Forensics Lab Speak to a forensic examiner, not a salesperson. INSTRUCTTHELAB NEWENQUIRIESEMAILE - DISCOVERY
32 pages · 53 min read
Identifying the right custodians is crucial for electronic disclosure.
45 pages · 86 min read
Early Case Assessment (ECA) is the early analysis of a dispute's facts, legal issues, and evidence, alongside Early Data Assessment (EDA) of electronically stored information.
Whether you are responding to a regulator, preparing for disclosure, or scoping an internal investigation, start the chain of custody with a short, confidential conversation.
