A privilege protecting documents and materials prepared by or for an attorney in anticipation of litigation from compelled disclosure to opposing parties.
Last reviewed: 2026/05/19
How attorney-client privilege applies when AI tools process confidential legal communications, and risks of inadvertent waiver through AI vendor data handling.
SecurityUsing AI to identify, notify custodians, and track preservation obligations when litigation or investigation triggers a duty to preserve electronically stored information.
SecurityThe intentional or inadvertent disclosure of privileged communications to a third party, potentially destroying attorney-client or work product protection.
Thomson Reuters' GPT-backed legal research and drafting with Westlaw integration (relaunched as CoCounsel Legal, 2025).
AI legal research pioneer (CARA AI); standalone retired 2025, its technology now powers Thomson Reuters CoCounsel.
Move from this definition to role-based legal AI shortlists and the selection criteria that matter for each type of legal team.
Am Law 200 and global firm workflows: accuracy at scale, security compliance, and matter-level auditability.
Last reviewed: 2026/05/19. Definitions are written by the LawyerAI Editorial team. Commercial relationships are disclosed and do not determine editorial scores or conclusions. See our Sponsorship & Affiliate Disclosure.
The work product doctrine, established by the Supreme Court in Hickman v. Taylor (1947) and codified in Federal Rule of Civil Procedure 26(b)(3), protects materials prepared by or for a party or its representative in anticipation of litigation from discovery by opposing parties. Unlike attorney-client privilege—which protects confidential communications between attorney and client—work product protection applies to the attorney's mental impressions, conclusions, opinions, and legal theories, as well as to factual material assembled in preparation for litigation regardless of whether it reflects legal analysis.
The doctrine recognizes two tiers of protection. Ordinary work product—factual compilations, interview summaries, document collections assembled for litigation—receives qualified protection that can be overcome by a showing of substantial need and inability to obtain the equivalent through other means. Opinion work product—the attorney's mental impressions, strategy, and legal theories—receives near-absolute protection that can be overcome only in exceptional circumstances.
In the AI context, work product doctrine questions arise when AI tools are used to generate litigation materials, analyze evidence, or develop legal strategy. Documents generated by AI tools at the direction of counsel—research memoranda, document review analyses, deposition outlines, damages models—may qualify as work product if created in anticipation of litigation. However, the extent to which AI-generated materials enjoy the same protections as attorney-generated materials, and under what circumstances AI system inputs and outputs constitute work product, are emerging questions that courts are only beginning to address.
Work product doctrine is fundamental to effective adversarial representation. Lawyers cannot effectively prepare a case if every strategic document, witness interview, and legal analysis is discoverable by opposing counsel. The doctrine creates the protected space in which attorneys develop their litigation theories and strategies without exposure.
AI's integration into litigation preparation raises novel work product questions. When a lawyer inputs case facts and asks an AI to identify legal theories or assess witness credibility, are the AI's outputs work product? When an AI analyzes opposing parties' documents and generates a strategy memorandum, does that memorandum receive protection? Existing doctrine does not clearly answer these questions, and the answers may depend on how much independent legal judgment the attorney contributed to directing the AI's work.
A practical concern is AI tool design: if AI platforms log user prompts and outputs in ways that are accessible to the vendor or stored on vendor infrastructure without adequate contractual protection, work product arguments may be weakened. Lawyers using AI for litigation preparation should evaluate whether their vendor relationships and data handling arrangements support work product characterization of AI-assisted materials.
Legal AI tools used in litigation contexts—CoCounsel, Casetext, Harvey—are generally structured with the work product doctrine in mind. These platforms operate under terms of service and data processing agreements that limit vendor access to customer work product, avoid using litigation-preparation materials for model training, and maintain confidentiality of user inputs and AI outputs.
Some platforms apply work product analysis directly: tools that assist with privilege review in e-discovery classify documents as ordinary versus opinion work product alongside attorney-client privilege designations, supporting privilege log preparation. AI-assisted legal research tools generate research memoranda that are typically treated as attorney work product when created in anticipation of litigation, subject to the same caveats about adequate human oversight and direction.
The practical risk area is inadvertent disclosure. Attorneys who share AI-generated litigation preparation materials with clients, consultants, or other third parties without considering work product implications may inadvertently waive protection. The same waiver analysis that applies to attorney-generated work product applies to AI-assisted materials—the key question is whether the disclosure is inconsistent with maintaining confidentiality against adversaries.