Legal Cyber Academy

Discovery of AI: Prompts, Training Data, and the New ESI

Taught by Daniel B. Garrie, Scot G. Doyen and 3 others

eDiscoveryLevel: Beginner58m

Overview of Discovery of AI: Prompts, Training Data, and the New ESI

This class examines how discovery practice applies to generative AI systems, including the prompts, outputs, and underlying model artifacts that may become relevant evidence in litigation. Faculty explain, in plain terms, what generative AI is and how it differs from traditional document retrieval, why outputs are mutable even when the same prompt is run twice, and how layered technology stacks (system prompts, fine-tuning, retrieval context, harnesses) affect what can later be reconstructed and produced. The discussion distinguishes inference-layer artifacts (prompts, outputs, system prompts and configuration, tool and agent logs, retrieval records, and usage telemetry) from model-layer artifacts (model cards, evaluation reports, version history, validation and testing records, and governance materials), and explains why some of these are far less complete than requesting parties expect. Panelists discuss why AI-related evidence can disappear quickly, including short vendor retention windows, contracts that do not give corporate customers access to underlying logs, model deprecation practices, and the growing use of consumer AI accounts for work-related tasks. The program also addresses possession, custody, and control questions under Rule 34 when AI tools or agents are involved, the practical difficulties of "deposing" an autonomous agent, and the importance of understanding an agent's memory and audit trail limitations. Practical guidance is offered on using interrogatories and Rule 26(f) conferences to narrow discovery requests, using the vendor's own terminology for artifacts, and working cooperatively with opposing counsel to focus requests given the technical complexity involved. Finally, the panel discusses privilege and work product considerations specific to AI tool use, including the risks of commingled expert use of AI tools and the practice of creating separate, logged tool instances for work done at counsel's direction, along with recent decisions addressing burden, proportionality, and the production of AI logs and expert prompts.

What you’ll learn in Discovery of AI: Prompts, Training Data, and the New ESI

  • Explain how generative AI produces outputs and why those outputs are mutable rather than fixed records
  • Identify inference-layer and model-layer artifacts that may exist and be subject to discovery requests
  • Apply existing federal discovery rules, including proportionality and burden standards, to AI-related requests
  • Assess possession, custody, and control issues raised by AI tools, agents, and consumer accounts
  • Draft interrogatories and Rule 26(f) positions that use accurate terminology to target discoverable AI artifacts
  • Recognize privilege and work product risks arising from attorney, client, and expert use of AI tools

Audience and prerequisites

Who should take this course — Discovery of AI: Prompts, Training Data, and the New ESI

Litigators and in-house counsel handling or anticipating discovery disputes involving AI tools, and technologists or e-discovery professionals who support legal teams on AI-related evidence issues.

Prerequisites for Discovery of AI: Prompts, Training Data, and the New ESI

None.

Curriculum

  1. 1. Introducing AI discovery: speakers, disclaimer, and webinar agenda

    Watch Introducing AI discovery: speakers, disclaimer, and webinar agenda
  2. 2. What generative AI actually is and how it produces mutable outputs

    Watch What generative AI actually is and how it produces mutable outputs
  3. 3. AI vocabulary explained: prompts, training data, and model layers

    Watch AI vocabulary explained: prompts, training data, and model layers
  4. 4. The governing principle: applying old discovery rules to new AI data

    Watch The governing principle: applying old discovery rules to new AI data
  5. 5. What AI evidence actually exists: inference layer and model layer artifacts

    Watch What AI evidence actually exists: inference layer and model layer artifacts
  6. 6. Why AI evidence disappears before you ever request it in discovery

    Watch Why AI evidence disappears before you ever request it in discovery
  7. 7. Who possesses AI data, and what happens when the AI itself is the actor

    Watch Who possesses AI data, and what happens when the AI itself is the actor
  8. 8. Building effective discovery requests for AI artifacts using interrogatories and Rule 26(f)

    Watch Building effective discovery requests for AI artifacts using interrogatories and Rule 26(f)
  9. 9. Privilege, work product protection, and recent AI discovery case law

    Watch Privilege, work product protection, and recent AI discovery case law

Slides & materials

Your instructors

Portrait of Daniel B. Garrie

Daniel B. Garrie

Founder, Law & Forensics; Neutral at JAMS; Faculty at Harvard

Daniel B. Garrie is the founder and executive managing partner of Law & Forensics LLC, a boutique cybersecurity and forensic engineering firm he co-founded in 2008. He serves as a neutral, arbitrator, and forensic special master at JAMS, focusing on cybersecurity, cryptocurrency, and complex technology disputes, and holds faculty appointments at Harvard and Rutgers Law School. A patented software inventor and prolific author, he is editor-in-chief of the Journal of Law & Cyber Warfare and a widely cited authority on computer forensics, eDiscovery, and cyber litigation.

Scot G. Doyen

Shareholder, Clausen Miller P.C.

Scot G. Doyen is a shareholder in the Houston office of Clausen Miller P.C. He has about 30 years of trial experience in commercial, personal injury and first-party insurance matters, and appellate experience that includes briefing before the U.S. Supreme Court and arguments before the Texas Supreme Court and the Fifth Circuit. He has served in 14 MDLs, usually as liaison counsel, and is Board Certified in Civil Trial Law by the Texas Board of Legal Specialization.

Karen Silverman

CEO and Founder, The Cantellus Group

Karen Silverman is CEO and founder of The Cantellus Group, which advises Fortune 50 companies, startups, consortia, and public agencies on AI strategy, risk management, and readiness for successful adoption. Before founding the firm she spent more than two decades in practice and leadership at Latham & Watkins LLP, working on complex antitrust matters, M&A, governance, and crisis management, and led Latham's partnership with the World Economic Forum on its Board Toolkit for Artificial Intelligence; she retired as a Latham partner in March 2020. She serves on the ABA Task Force on the Law and Artificial Intelligence and on the boards of AIEDU and Not For Sale, and holds a B.A. from Wellesley College and a J.D. from UC Berkeley School of Law. At Legal Cyber Academy, Karen Silverman teaches on navigating AI and its ethical challenges and on AI regulatory efforts in the U.S. and the E.U.

Mari Grace

Counsel, Debevoise & Plimpton LLP

Mari Grace is counsel in the Data Strategy & Security group at Debevoise & Plimpton LLP in New York. She advises clients on AI governance, regulatory compliance and litigation risk, including assessing AI use cases, building governance frameworks and negotiating AI-related agreements. She has more than a decade of litigation experience in antitrust, securities class actions, derivative suits and M&A disputes. She is admitted in New York and the District of Columbia.

Patricia Antezana

Counsel, Reed Smith LLP

Patricia E. Antezana is counsel at Reed Smith LLP in Pittsburgh, working in the firm's litigation and e-discovery practices. She manages electronic discovery across the litigation lifecycle, advising on legal holds and preservation, early case assessment, ESI protocols, collections, document review, privilege issues and productions, and has run large-scale review projects in complex litigation. Her wider litigation background covers commercial disputes, business torts and product liability defence. She holds a J.D. from Boston College Law School and an A.B. in government, cum laude, from Harvard College, and is admitted in Pennsylvania. At Legal Cyber Academy she teaches on artificial intelligence and cloud computing in eDiscovery.