Documents that a party’s romantic partner created with ChatGPT in anticipation of litigation were not protected by the work product privilege, a Massachusetts trial court judge has found.

The plaintiff was embroiled in a dispute over whether he was in default of a $2 million promissory note. He emailed the loan documents at issue and a letter claiming default of the loan to his girlfriend for her to run through ChatGPT. She then emailed him a draft response letter and other documents created by ChatGPT. The plaintiff’s lawyer was not involved.

Once litigation commenced, the defendants moved to compel production of the email communications between the plaintiff and his girlfriend, including the ChatGPT documents.

The plaintiff argued in response that the documents qualified for work product protection because they were prepared in anticipation of litigation.

But Judge Debra A. Squires-Lee disagreed.

The judge distinguished the scenario from Morgan v. V2X, Inc., a 2026 case in which the U.S. District Court in Colorado held that materials that a pro se plaintiff prepared by AI could be subject to the work product privilege. Squires-Lee noted that the plaintiff was represented by sophisticated counsel who had nothing to do with his decision to ask his romantic partner to upload the materials to ChatGPT and generate information and documents.

“[This case is] more akin to United States v. Heppner, [a 2026 case] in which the [U.S. District Court for the Southern District of New York] held that AI documents generated by a represented Defendant’s use of AI on his own ‘volition’ ‘do not merit protection under the work product doctrine’ because they were not prepared by or at the ‘behest’ of counsel and did not reflect ‘counsel’s strategy,’” the judge wrote. “I conclude, as a matter of first impression, that 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.”

The case is Shealy, et al. v. Seaside Investments, et al.

Foley Hoag cyberlaw attorney Colin J. Zick said he was pleased to see the judge follow Heppner, which he described as quickly becoming the leading case on the issue.

“The work product doctrine exists to protect the adversarial process by shielding the work of legal advocates,” Zick said. “When a party circumvents that process entirely, whether personally as in Heppner or through a romantic partner as in Shealy, the rationale for protection does not exist.”

Kurt B. Fliegauf, a civil litigator at Conn Kavanaugh, noted a divergence between Heppner and Morgan.

Here, Fliegauf said, Squires-Lee squared those two decisions by distinguishing between a pro se litigant, who may need AI to develop the legal theories of their case, and a party represented by counsel, who does not have the same need.

Given that, lawyers should continue to warn clients not to use AI to develop legal strategies, he said.

“This is not just to limit a client’s second-guessing of the attorney’s strategy,” Fliegauf said. “The prompts that a client inputs into AI could constitute a damning admission and most likely is not protected.”