Artificial Intelligence is quickly being woven into our everyday lives. From drafting wedding speeches to creating vacation itineraries, many people turn to generative AI on a daily basis. The same is true in the legal world, but as use of generative AI increases, so do the risks. As this alert explains, those risks include not only made-up legal authorities but forced disclosure of confidential information.
You have likely heard about lawyers being sanctioned for referencing non-existent cases or “hallucinations” in their pleadings. Massachusetts is not exempt from this trend. In July 2025, a Middlesex County Superior Court Judge sanctioned an attorney for her use of ChatGPT resulting in her citation not only to hallucinated cases, but to non-existent quotes and court holdings in a real case. In her sanctions order, the Judge noted that “Lawyers have a duty to provide competent representation of their clients and demonstrate candor towards the tribunal. … This extends to submitting filings that accurately state the law and present arguments that supports their clients’ interests. By submitting motions and briefs that contain hallucinations and misstatements of the law, the lawyer violates their professional obligations.” Dastou v. Holmes (Lawyers Weekly No. 12-026-25) (3 pages) (Goldenberg, J.) (Middlesex Superior Court) (Civil Action No. 2381CV02212).
The use of AI in litigation does not end with attorneys. Individuals are turning to generative AI themselves for their legal matters, at their peril. One recent Massachusetts case demonstrates both the advantages of hiring savvy counsel and the risks of helping yourself. In Shealy, et al. v. Seaside Investments, et al. the Plaintiff received legal documents from the Defendant and shared them with his girlfriend. The girlfriend uploaded the documents to ChatGPT and asked it to draft a legal letter in response, which she then emailed to the Plaintiff, along with other documents created by ChatGPT. No attorneys were involved in the preparation of the letter or documents created by ChatGPT. During the course of the ligation the Defendant asked for the letter and documents, and the Plaintiff sought to withhold them under the legal protection for attorney work product.
In order to be protected as “work product,” a document must be prepared in anticipation of litigation and, importantly, prepared by or for the party or their representatives. In Shealy, the court considered whether the girlfriend or ChatGPT could be considered the Plaintiff’s representatives for the purposes of work product. The Court refused to expand the definition of representative to include a romantic partner who had no other connection to the litigation. In considering whether ChatGPT would be considered a ‘representative,’ the court relied on Morgan v v2X, Inc, U.S. Dist. Ct., No. 25–CV–01991–SKC–MDB, in which a U.S District Judge in Colorado held that a pro se litigant (someone representing themselves), who is both the party and their own advocate, may seek work product protection over their use of generative AI. In comparing the Shealy matter to the Morgan matter, however, the court emphasized that the Plaintiff in Shealy was represented by counsel, who had nothing to do with Plaintiff’s decision to have his girlfriend upload the documents to ChatGPT and create a draft legal response. More pointedly, the court stated that “AI is a tool not a person.”
It is common for clients to share their frustrations, setbacks and even legal strategies with family members or friends. However, if you do so in writing, whether in the form of a text message, signal chat or email, those documents may be subject to compelled production by the other side in litigation. As a family law attorney, I regularly remind my clients not to text or email with their parents, new love interests or friends about their legal cases for this very reason. Savvy counsel can and will ask for these documents.
More pointedly, as you begin to think about litigation or if you are already embroiled in it, use of generative AI searches without the assistance of counsel should be avoided. Some of the commonly used AI tools may be misleading, containing inaccurate information and/or hallucinations; experienced attorneys know which tools are reliable. More importantly, the prompt questions and results may have to be disclosed to the other side, putting your entire case at risk. The questions and results can provide a glimpse into your mindset and strategic thoughts or reveal information you would not otherwise share with the opposing side.
Don’t take those unnecessary risks. The attorneys at Prince Lobel are here to support you and answer all of your questions, big or small. Our advice comes with years of training and experience and is protected from compelled disclosure unlike legal advice from your friends, family, or ChatGPT.