In its June 16, 2026, decision in Earl Wade Shealy Jr. v. Seaside Investments, LLC, No. 2684CV00799-BLS2, the Suffolk County Superior Court’s Business Litigation Session (Squires-Lee, J.) compelled production of documents that the plaintiff’s romantic partner had generated using ChatGPT. The court held that, on the facts before it, neither AI queries for legal assistance nor the resulting output were protected work product when made by someone who is not a party’s representative and not at the direction of counsel—even though the party was represented and the materials were created in anticipation of litigation. Although the trial court’s order is not binding on other courts, the decision offers an early and instructive window into how judges may treat AI-assisted materials in discovery.
Background
The dispute arose from alleged breaches of an agreement on the purchase and sale of a gas station in Boston’s South End neighborhood. The plaintiff, Earl Wade Shealy, Jr., alleges that the defendants defaulted on a promissory note and loan agreement by failing to pay as required under the agreements’ terms. Seaside Investments LLC responded by sending a letter to Shealy also alleging default. Shealy, in turn, forwarded the document to his partner, Debbi Fields.
Fields then uploaded the documents to ChatGPT, generated a response letter and other documents using the AI tool, and emailed the output back to Shealy. No attorney was involved in Fields’s use of ChatGPT. Fields testified she was merely “like a Xerox,” assisting with “providing a paperwork flow.”
Defendants moved to compel production of (i) the ChatGPT communications and output, and (ii) documents relating to Shealy’s financial ability to prepay the note. Plaintiffs argued that the AI-generated documents were prepared in anticipation of litigation and therefore were protected under the work product doctrine.
The court’s analysis: AI output and the work product doctrine
The court’s decision turned on an analysis of three major requirements for the work product doctrine, which protects (1) documents and tangible things (2) prepared by or for another party or by or for that party’s representative (3) in anticipation of litigation or for trial.
A non-attorney third party is not a “representative”
The court held that Fields, as Shealy’s romantic partner, was not a “representative” under the work product doctrine. The types of persons whose documents are protected under Mass. R. Civ. P. 26(b)(3), including a party’s “attorney, consultant, surety, indemnitor, insurer, or agent,” do not include a friend or partner with “absolutely no substantive connection to the litigation or to plaintiffs’ counsel.” Expanding the definition to cover such individuals, the court reasoned, would not “enhance the vitality of an adversary system” and would “expand the work product doctrine without meaningful limitation.”
ChatGPT is not a “representative” either, unless used at counsel’s direction
Plaintiffs cited two recent federal decisions—Morgan v. V2X, Inc., No. 25–CV–01991–SKC–MDB, at *2 (D. Colo. Mar. 30, 2026) and Warner v. Gilbarco, Inc., 820 F. Supp. 3d 629, 636 (E.D. Mich. 2026)—both of which held AI-generated materials are subject to work product protection. The court distinguished those cases because both involved pro se litigants using AI as both party and advocate. Here, plaintiffs were represented by “sophisticated, capable counsel.”
The court found the case more akin to United States v. Heppner, 820 F. Supp. 3d 292 (S.D.N.Y. 2026), in which AI documents generated by a represented defendant on his own “volition” did not merit work product protection because they were not prepared at the “behest” of counsel and did not reflect “counsel’s strategy.”
The court’s holding on this issue of first impression:
Neither the queries of AI programs for legal assistance by a party’s romantic partner nor the AI output are protected from disclosure under the work product doctrine unless performed at the direction of counsel.
AI output is not opinion work product
The court further held that ChatGPT’s output does not qualify as opinion work product, which is discoverable only in extreme circumstances. As the court stated: “AI is a tool, not a person. Nothing about what ChatGPT produces discloses the ‘mental impressions, conclusions, opinions, or legal theories of an attorney or other representative of a party concerning the litigation.’”
Practical takeaways for companies and individuals
While Massachusetts courts have already confronted AI—most notably in sanctioning attorneys for citing fabricated, AI-generated cases (such as Judge Brian A. Davis’ 2024 warning to the “broader bar” in Smith v. Farwell, No. 2282CV01197 (Mass. Super. Ct. Feb. 12, 2024))—Shealy appears to be among the first to reach AI through the work product doctrine, and one of the earliest anywhere to do so involving a represented party. The court’s reasoning sends a strong signal: the work product shield may not extend to AI use by someone who is not a party’s “representative” and who is acting outside counsel’s oversight. Companies[1] and individuals would be well served to keep the following in mind:
Route litigation-related AI use through counsel. The court’s favorable citation to United States v. Heppner is a strong signal that work product protection may turn on whether AI is used at the “direction of counsel.” Parties are well advised to involve counsel before using AI tools to draft or analyze litigation-related documents. Under the court’s reasoning, even a party’s own, self-initiated AI use could fall outside the doctrine’s protection.
Think twice before sharing litigation documents with non-attorneys for AI analysis. Under Shealy, sending legal documents to a spouse, partner, family member, or friend who then uses AI to generate analysis, or draft responses may carry no privilege or work product protection. A close family relationship may not change the analysis: the marital privilege protects confidential communications between spouses, and the court’s reasoning suggests it may not reach documents a spouse creates by feeding materials into a public AI tool. Similar questions could arise as to parents, children, and other relatives, and the risk that such materials are ordered produced is a real one.
Address AI use during disputes in internal policies. Organizations are well positioned to get ahead of this issue by making clear when employees should consult counsel before using AI tools to analyze or respond to legal matters, including demand letters, regulatory inquiries, and contracts in dispute.
Revisit litigation hold and preservation procedures. Companies are well served to look at whether their litigation hold notices reach AI-generated content, and whether employees understand any obligations to preserve AI interactions relating to disputes, since opposing parties may seek their production.
Favor counsel-directed or enterprise AI tools for litigation work. The decision is not a reason to avoid AI, but a strong reason to be deliberate about how it is used. Where counsel directs the AI use, or where the work runs through an enterprise platform with appropriate confidentiality and data-handling protections rather than a public chatbot, AI-assisted materials stand a far better chance of qualifying as protected work product. Counsel can help structure AI use so that its benefits are not offset by discoverability risk.
- Although Shealy is a Massachusetts trial court decision that is not binding elsewhere, and some courts-particularly in cases involving pro se litigants-have reached the opposite result and extended work product protection to party-initiated AI use, the work product principles Shealy applies are not unique to Massachusetts and have already been echoed in other jurisdictions. Because other courts may adopt the same reasoning, these takeaways are prudent for companies and individuals in any jurisdiction, who should also consult counsel regarding the state of the law where they operate.
[Back to reference]