Artificial intelligence is transforming modern litigation into ways that few anticipated even a decade ago. Courts, litigators, and corporate legal departments now operate in an environment where vast volumes of electronically stored information are generated every day, much of it created, curated, or filtered by AI-driven systems. From predictive coding in document review to automated communication tools that produce transient data, artificial intelligence has become intertwined with the lifecycle of evidence. As a result, discovery obligations under existing procedural frameworks must be reinterpreted through the lens of emerging technologies. Lawyers and organizations alike face a pressing question: how can they responsibly manage AI’s influence on data creation, retention, and retrieval while meeting their duties under the law?
The expansion of electronically stored information has already reshaped litigation practice. Email, messaging platforms, collaborative workspaces, cloud storage, and enterprise software produce enormous data trails. Artificial intelligence compounds this reality by enabling automated content generation, smart categorization, and algorithmic decision-making. AI systems do not merely store data; they transform it, summarize it, and in some cases create entirely new content based on user prompts or training datasets. The evidentiary landscape is therefore no longer confined to static documents or clearly identifiable communications. Instead, it includes dynamic outputs generated by models whose internal processes may be opaque even to their developers. This evolution demands renewed scrutiny of how discovery obligations are interpreted and enforced under the Federal Rules of Civil Procedure, particularly Rules 26 and 34, which define the scope and mechanics of discovery.[1]
Under Rule 26, parties may obtain discovery regarding any nonprivileged matter that is relevant to a claim or defense and proportional to the needs of the case.[1] Rule 34 governs requests for production of electronically stored information, while Rule 37(e) addresses failures to preserve such information.[1] These provisions were intentionally drafted to remain technologically neutral, yet they now must accommodate generative AI systems and machine learning architectures that produce and modify information at scale. Courts increasingly confront disputes involving algorithmic outputs, automated decision logs, and ephemeral data generated by AI-enhanced platforms. Although the rules themselves remain constant, their application becomes more nuanced when evidence is shaped or created by artificial intelligence.
One of the most significant ways AI intersects with litigation is through technology-assisted review, commonly known as predictive coding. Courts have recognized that machine learning tools can be used to identify relevant documents in large data sets, provided that the process is reasonable and defensible. In Da Silva Moore v. Publicis Groupe, a federal court endorsed the use of predictive coding in civil discovery, signaling judicial acceptance of AI-assisted review when implemented transparently and in good faith.[2] Since that decision, technology-assisted review has become more sophisticated, incorporating advanced analytics and iterative training protocols. Nevertheless, litigants who rely on such tools must be prepared to explain how the system was trained, validated, and monitored. Transparency and cooperation remain central to judicial approval.
The advantages of AI-driven review are substantial. Advanced systems can identify conceptual relationships across millions of documents, reduce duplication, and accelerate production timelines. In complex commercial litigation, these efficiencies can significantly reduce cost and delay. Yet reliance on AI does not eliminate responsibility. If a producing party fails to take reasonable steps to ensure accuracy and completeness, it risks violating discovery obligations under Rule 26’s proportionality framework.[1] Courts expect parties to document their methodologies and demonstrate that their processes align with best practices articulated in resources such as The Sedona Principles, which emphasize reasonableness and cooperation in electronic discovery.[5]
Beyond document review, AI increasingly contributes to the substantive creation of business records. Generative AI tools draft correspondence, generate reports, and produce analytical summaries that may later become central evidence in litigation. When such materials are introduced in court, questions arise regarding authorship, authenticity, and intent. Federal Rule of Evidence 901 requires that evidence be authenticated as a condition of admissibility.[3] In cases involving AI-generated content, authentication may require evidence of system integrity, user prompts, and audit logs. The involvement of AI does not negate evidentiary standards; rather, it adds additional layers that litigants must address to satisfy authenticity requirements.
The duty to preserve relevant evidence arises when litigation is reasonably anticipated. Once that duty attaches, organizations must implement litigation holds that encompass all relevant sources of electronically stored information.[1] In an AI-integrated environment, identifying those sources requires technical literacy and collaboration between legal and information technology teams. AI systems often generate metadata, logs, prompt histories, and model outputs that may be relevant to claims or defenses. If such information is lost due to automatic deletion or inadequate retention policies, Rule 37(e) authorizes courts to impose measures to cure prejudice or, in cases of intent to deprive, more severe sanctions.[1] The existence of automated systems does not excuse a failure to preserve; courts focus on whether reasonable steps were taken.
Ephemeral communications present particular challenges in this context. Messaging platforms that allow automatic deletion or disappearing messages are widely used in business operations. When AI tools integrate with these platforms, such as by summarizing or auto-generating responses, the resulting data may be transient unless proactively preserved. Courts have increasingly scrutinized the use of ephemeral messaging in situations where litigation was foreseeable. Although such technologies are not inherently improper, parties must ensure that auto-deletion features are suspended once preservation duties arise. Failure to do so may be viewed as a failure to take reasonable preservation steps under Rule 37(e).[1]
AI systems also raise questions regarding explainability and expert testimony. In disputes where algorithmic decisions are central to the case, parties may seek to introduce expert analysis explaining how a model functions and whether its outputs are reliable. The admissibility of such testimony is governed by the standards set forth in Daubert v. Merrell Dow Pharmaceuticals, Inc., which require that expert opinions be based on reliable principles and methods applied to sufficient facts.[4] As AI technologies become more complex, courts must evaluate whether proposed expert testimony adequately addresses issues such as model validation, bias, and error rates. Litigants who fail to establish methodological reliability risk exclusion of critical evidence.
Privilege considerations further complicate AI integration in litigation. When attorneys use AI tools to draft documents or analyze case materials, the confidentiality of client information must be preserved. Disclosure of privileged information to third-party vendors may, under certain circumstances, jeopardize the protection afforded by attorney-client privilege. Careful review of vendor agreements and data handling policies is therefore essential. Although not every interaction with an AI system constitutes waiver, prudent counsel must ensure that safeguards align with professional obligations of competence and confidentiality.
The global nature of AI infrastructure introduces additional discovery complexities. Cloud-based systems may store data across multiple jurisdictions, implicating foreign data protection laws. When cross-border discovery arises, courts must balance domestic discovery obligations with principles of international comity. Although the Federal Rules emphasize broad access to relevant evidence, parties may face restrictions under foreign statutes. Effective litigation strategy requires early assessment of data location, applicable privacy regimes, and potential conflicts of law.
Cost and proportionality remain central themes in AI-related discovery disputes. Rule 26 underscores that discovery must be proportional to the needs of the case, considering factors such as the importance of the issues, the amount in controversy, and the burden of proposed discovery.[1] AI can reduce review burdens, but disputes over algorithmic transparency or production of training data may increase complexity. Courts expect parties to articulate specific burdens rather than rely on conclusory objections. Reference to widely recognized best practices, including The Sedona Principles, may assist courts in evaluating reasonableness.[5]
Spoliation sanctions represent one of the most significant risks in AI-driven environments. If electronically stored information that should have been preserved is lost because a party failed to take reasonable steps, courts may order measures to cure prejudice or impose more severe sanctions upon finding intent to deprive another party of the information’s use in litigation.[1] AI systems that overwrite logs or delete transient data must be configured in a manner consistent with preservation duties. Documentation of retention policies and litigation hold procedures is often decisive in determining whether a party acted reasonably.
The authenticity of AI-generated evidence also demands careful attention. As generative technologies evolve, concerns about fabricated or manipulated content increase. Federal Rule of Evidence 901 requires sufficient evidence to support a finding that the item is what the proponent claims it is.[3] In cases involving AI-generated images, text, or audio, parties may need to present technical evidence establishing system reliability and chain of custody. Challenges to authenticity are likely to grow as deep-fake technologies become more sophisticated.
Judicial perspectives on AI continue to evolve. Early cases endorsing predictive coding signaled a willingness to embrace innovation when implemented responsibly.[2] At the same time, courts emphasize cooperation, transparency, and proportionality in discovery practice, as reflected in The Sedona Principles.[5] As AI becomes more embedded in litigation workflows, standards of reasonableness will likely develop through case law and judicial guidance. Attorneys who remain informed about technological developments and judicial expectations will be better positioned to navigate disputes effectively.
Ultimately, the integration of artificial intelligence into litigation practice reflects a broader transformation of the information ecosystem. AI offers tools that can streamline discovery, enhance analytical precision, and manage large data volumes. Yet it also introduces new preservation challenges, evidentiary complexities, and ethical considerations. The foundational principles of fairness, transparency, and diligence remain constant. By aligning AI practices with the Federal Rules of Civil Procedure,[1] adhering to evidentiary standards under the Federal Rules of Evidence,[3] satisfying reliability requirements articulated in Daubert,[4] drawing upon judicial guidance in Da Silva Moore,[2] and following best practices articulated by The Sedona Conference,[5] litigants can responsibly navigate the evolving terrain of AI-driven litigation.
Contact Tishkoff
Tishkoff PLC specializes in business law and litigation. For inquiries, contact us at www.tish.law/contact/. & check out Tishkoff PLC’s Website (www.Tish.Law/), eBooks (www.Tish.Law/e-books), Blogs (www.Tish.Law/blog) and References (www.Tish.Law/resources).
Sources
- Federal Rules of Civil Procedure, Rules 26, 34, and 37(e). https://www.federalrulesofcivilprocedure.org/frcp/title-v-disclosures-and-discovery/
- Da Silva Moore v. Publicis Groupe, 287 F.R.D. 182 (S.D.N.Y. 2012). https://law.justia.com/cases/federal/district-courts/new-york/nysdce/1:2011cv01279/375665/96/
- Federal Rules of Evidence, Rule 901. https://www.law.cornell.edu/rules/fre/rule_901
- Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993). https://supreme.justia.com/cases/federal/us/509/579/
- The Sedona Conference, The Sedona Principles, Third Edition: Best Practices, Recommendations & Principles for Addressing Electronic Document Production (2018). https://www.thesedonaconference.org/publication/The_Sedona_Principles
This publication is for general informational purposes and does not constitute legal advice. Reading it does not create an attorney-client relationship. You should consult counsel for advice on your specific circumstances.