Evolving Discovery Strategies for the AI Era

Most discussions about generative AI and discovery focus on what organizations must preserve, collect, and produce. But discovery has always been a two-way street. As plaintiffs and other litigants increasingly use generative AI to draft resumes, summarize meetings, prepare declarations, calculate damages, revise communications, and organize factual narratives, those AI interactions may themselves become an important potential source of electronically stored information.

Prompts and chats may provide evidence that former employees used their former employer’s confidential information to generate work for a new employer. AI conversations could also reveal employees asking AI whether certain actions would violate employment contracts or confidentiality provisions. In appropriate cases, looking beyond finished documents may provide meaningful insight into how key evidence was created, revised, or framed.

Effective litigators have long recognized that discovery is about more than collecting documents. It is also an opportunity to develop facts, test claims, and position cases for successful resolution. As AI becomes embedded in everyday life, litigants should consider whether discovery strategies should evolve to account for this new category of potentially relevant evidence.

Key Takeaways

• AI use may create potentially discoverable electronically stored information (ESI) that provides context beyond the final document.
• Discovery obligations apply equally to all parties, making AI-generated materials a potential source of relevant evidence from opposing litigants.
• AI-related discovery should remain targeted, proportional, and tied to the claims and defenses at issue.
• Early consideration of AI use during case planning and discovery discussions may help avoid unnecessary disputes.
• Litigants who recognize where AI-related evidence may exist may be better positioned to develop a complete factual record.

DISCOVERY RUNS BOTH WAYS

Discovery is often viewed through a defensive lens, with organizations focused on their own preservation and production obligations. Those obligations remain important, particularly as generative AI becomes integrated into routine business operations. But they are not limited to defendants.

Plaintiffs and other litigants are likewise responsible for preserving and producing relevant, nonprivileged AI-generated materials. As a result, discovery presents an opportunity to affirmatively develop the factual record by seeking such AI-generated evidence from the opposing party where appropriate.

Experienced litigators routinely seek emails, text messages, metadata, revision histories, and other forms of ESI that reveal how information evolved over time. Generative AI introduces another potential source of evidence that may provide insight into how key documents, analyses, and communications were developed. Depending on the circumstances, AI interactions may provide additional context regarding how documents were developed, what information was considered, how narratives evolved, or how analyses were performed.

This does not mean that every AI interaction will be relevant or discoverable. Like all ESI, AI-generated materials remain subject to the familiar principles of relevance, proportionality, privilege, and work-product protection. However, as AI becomes a routine part of how individuals create and revise information, litigants should consider whether those materials warrant attention during discovery.

LOOKING BEYOND THE FINAL DOCUMENT

The final version of a document rarely tells the entire story. This has long been true of emails, tracked changes, spreadsheets, and collaborative workspaces. AI-assisted drafting introduces another layer to the document creation process that, under appropriate circumstances, may provide useful context regarding the finished product.

Consider an employment dispute involving an employee’s hiring or job qualifications. A resume, job application, or cover letter may have been prepared with assistance from a generative AI tool. The final document itself may be responsive, but information concerning how it was developed may also be relevant if it bears on disputed issues in the case.

Similarly, parties increasingly use AI tools to summarize information, refine such summaries, organize factual narratives, perform calculations, or assist with drafting documents that later become relevant in litigation. Depending on the circumstances, understanding how those materials were developed may provide additional insight into their accuracy, completeness, or reliability. AI chats may show less favorable information was purposefully removed from summaries or factual narratives, while drafting assistance may have been used to obfuscate the use of an employer’s information in creating materials for a new business.

The point is not that AI-generated materials should become the focus of every discovery request. Rather, litigants should recognize that AI use may create evidence that did not exist only a few years ago. Overlooking those materials may mean overlooking relevant facts.

DISCOVERY STRATEGIES SHOULD EVOLVE ALONGSIDE TECHNOLOGY

Traditional discovery requests were developed before generative AI became part of routine life. As a result, parties cannot always assume that existing discovery strategies or older request templates adequately account for AI use. Yet, the drafters of the modern Federal Rules recognized that technology would continue to evolve and declined to define ESI narrowly, allowing the term to encompass new data types.

Accordingly, AI use may warrant consideration during early case planning, preservation discussions, and meet-and-confer conferences where appropriate. Discovery requests should continue to be tailored to the issues in dispute, but they should also reflect the reality that relevant information may now exist in formats and locations that were uncommon only a few years ago. Rule 26(f) conferences, interrogatories, and deposition questions may seek information regarding a plaintiff’s use of AI tools.

The goal is not to seek every AI interaction that may exist. Broad, unfocused requests are unlikely to satisfy proportionality requirements and may generate unnecessary disputes.

Instead, the discovery strategy should remain grounded in the facts of the case. When AI-assisted content creation intersects with the claims and defenses at issue, litigants should thoughtfully evaluate whether additional discovery may be appropriate.

QUESTIONS LITIGANTS SHOULD BE ASKING EARLY IN DISCOVERY

As generative AI becomes part of everyday business and personal life, litigants should consider whether AI prompts, chats, logs, and outputs may be relevant to the claims and defenses at issue. Addressing these issues early can help shape preservation efforts, inform discovery strategy, and reduce unnecessary disputes later in the litigation.

• Parties should consider the following questions: Could generative AI have been used to create or revise documents that are central to the claims or defenses? AI-assisted drafting is increasingly common across a wide range of business and personal activities and may warrant consideration when evaluating potentially relevant evidence.
• Could AI-generated materials provide useful context regarding how key documents or communications were developed? In some matters, AI interactions may offer insight into the evolution of evidence beyond what appears in the final document.
• Should AI use be addressed during early case planning and meet-and-confer discussions? Raising the issue early, where appropriate, may help clarify expectations regarding preservation and the scope of discovery.
• Are discovery requests appropriately tailored to the issues in dispute? AI-generated materials should be evaluated using the same principles that govern all discovery, including relevance, proportionality, and privilege.

THE BOTTOM LINE

Much of the discussion surrounding generative AI has focused on the new discovery obligations organizations face. Those obligations remain important, but they are only part of the story.

As generative AI becomes a routine part of how individuals create, revise, and communicate information, AI prompts, chats, logs, and outputs may also become increasingly important sources of potentially relevant evidence. Litigants who look beyond the finished document and thoughtfully consider how AI may have shaped key evidence will be better positioned to develop a complete factual record while remaining grounded in the traditional principles that have long governed discovery.