Is Your Company’s AI Use Discoverable? Work Product Privilege in the Age of AI

Artificial intelligence (AI) has quickly become part of how businesses operate and how their disputes are litigated. As companies increasingly turn to AI tools to research, draft, and think through problems, a practical question follows them into the courtroom: if you use AI to prepare for litigation, can the other side force you to turn over your prompts and what the AI told you? Courts are only beginning to answer that question, and the early decisions have consequences for any business that relies on these tools.

The threshold question is whether a litigant’s AI use is protected by the work product privilege. So far, courts have answered yes—but only up to a point.

Work Product Privilege Protects AI Use, to an Extent

Generally, the work product doctrine protects from disclosure documents prepared in anticipation of litigation and the mental impressions, opinions, and theories of parties and their attorneys. However, a party waives this protection by disclosure to an adversary, or in circumstances that substantially increase the likelihood that an adversary will obtain the materials.

Recent court decisions have held that using AI to assist with civil litigation preparation does not waive work product protection because generative AI programs are tools, not persons.[1] While courts acknowledge that using AI technically “discloses” information to a third party because AI systems collect and store user data, the courts reason that this does not waive the work product privilege because it is highly unlikely that the information will fall into the hands of an adversary.[2] Thus, in both federal and Texas courts, parties may assert work product privilege to prevent the production of their AI conversations and outputs used in preparation for litigation.

However, parties must be mindful that the work product protection extends only so far.

The Limitations of Work Product Protection for AI Use

Although AI conversations and outputs may be protected as a party’s mental impressions or case strategy, the work product privilege is not a blanket shield against disclosure of every category of material related to a party’s AI use.

Foremost, AI use is still subject to longstanding legal principles governing the work product doctrine and must meet the definition of work product: material prepared or mental impressions developed in anticipation of litigation.

Further, the work product privilege does not prevent discovery of all information related to a party’s AI use. For example, parties may be required to disclose the name of the AI tool they are using and all discovery materials or products which the party has shared with AI.

Importantly, the work product privilege does not allow parties to share confidential information with AI platforms in violation of a protective order, nor shield parties from liability for such violation.

Why This Matters for Businesses

Many companies have been hesitant in adopting AI tools because of uncertainty over whether it would risk disclosure of their confidential information and trade secrets. These recent decisions provide greater confidence that companies can adopt AI tools while maintaining confidentiality without automatically exposing its AI use to an opponent during litigation. However, a business’s AI use may be exposed when it is self-initiated by an employee and does not reflect current legal strategy.

To preserve protection, businesses should put AI-governance guardrails in place before a dispute arises. Companies can prevent disclosure by using closed-system AI tools rather than public models, negotiating vendor terms so inputs are not retained or used to train the model, distinguishing between privileged legal AI use and ordinary business AI use, and training employees on appropriate AI usage.

Why This Matters for Practitioners and In-House Counsel

These decisions signal that courts are treating AI use as another tool in the lawyer’s toolkit, much like legal research databases or word processors. Lawyers should be encouraged to use AI to assist in preparing litigation strategy, but keep in mind the privilege’s limitations.

Attorneys should consider including AI use provisions in protective orders to prevent disclosure of confidential information and trade secrets. A well-drafted protective order should make it unquestionably clear whether, how, and to what extent confidential information may be shared with AI tools. One such provision proposed by a federal district court includes the following AI-specific language:

No party or authorized recipient may input, upload, or submit Confidential Information into any modern artificial intelligence platform, including any generative, analytical, or large language model-based tool (“AI”), unless the AI provider is contractually prohibited from: (1) storing or using inputs to train or improve its model; and (2) disclosing inputs to any third party except where such disclosure is essential to facilitating delivery of the service. Where disclosure to a third party is essential to service delivery, any such third party shall be bound by obligations no less protective than those required by this Order. In addition, the AI provider must contractually afford the party or authorized recipient the ability to remove or delete all Confidential information upon request. A party intending to use AI that it contends meets these requirements must retain written documentation of these contractual protections.[3]

Practically speaking, this provision protects against the risks associated with mainstream AI tools that persistently collect and store data and could compromise confidentiality while simultaneously permitting parties to reap the benefits of using modern AI tools in the many ways which do not involve uploading confidential information.

BoyarMiller drafts protective orders with AI-specific language tailored to the practical realities of AI systems and to your particular dispute, putting the right guardrails in place to protect your confidential information. No matter how sophisticated or trustworthy an AI tool may appear, businesses need a protective order that keeps confidential information off any platform lacking the contractual safeguards to protect it.

If your company uses AI or expects to litigate against someone who does, our litigation team can help you understand your exposure and put the right protections in place. Contact BoyarMiller to discuss how these developments affect your business.

[1] See Warner v. Gilbarco, Inc., 820 F. Supp. 3d 629, 636 (E.D. Mich. 2026); Morgan v. V2X, Inc., No. 25–CV–01991–SKC–MDB, 2026 WL 864223, at *5 (D. Colo. Mar. 30, 2026); Court Minute Entry Regarding ChatGPT Materials In Camera Review, Tate Group Automotive, LLC v. Legacy Automotive Capital, LLC et al., Cause No. 25-BC11B-0020, Tex. Bus. (11th Div. Jun. 3, 2026).

[2] Morgan, 2026 WL 864223, at *5.

[3] Morgan, 2026 WL 864223, at *7.