Summary:
- Development: Three recent federal rulings (Heppner, Warner, Morgan) produced differing results but collectively signal courts are examining AI chat logs in discovery.
- Why it matters: Pennsylvania family cases often involve self-represented parties and highly sensitive facts, increasing the risk that AI chats could be requested in discovery and harm privilege or strategy.
- Action point: Avoid sharing case-specific facts or attorney advice with consumer AI, use hypotheticals, review platform privacy settings or enterprise options, and consult counsel before using AI related to your case.
If you have ever typed legal questions into ChatGPT or Claude looking for answers about Pennsylvania divorce, child custody, child support, or what to expect in family court, you are not alone. These AI tools are presented as fast, free, and readily available. However, recent court decisions suggest that what you share with an AI chatbot may not stay private, and in a Pennsylvania family law case, that can matter.
Recent court decisions have required parties to produce AI chat logs and chatbot-generated content during discovery. As a result, individuals involved in Pennsylvania divorce, custody, or support matters should not assume that conversations with platforms such as ChatGPT or Claude are private or protected from disclosure. Especially in family law disputes where issues often involve children, finances, parenting, and personal credibility, sharing sensitive information with consumer AI tools can create unnecessary risk. Before using AI to seek legal guidance or discuss case-related matters, it is important to consider how recent rulings, including United States v. Heppner (S.D.N.Y. Feb. 17, 2026) and Warner v. Gilbarco, Inc. (E.D. Mich. Feb. 10, 2026), may influence discovery obligations and litigation strategy in Pennsylvania family law cases.
Can AI Chat Logs Be Discovered in Court?
Three federal decisions issued in early 2026 are shaping how courts across the country are approaching AI chat logs in litigation.
United States v. Heppner (S.D.N.Y. Feb. 17, 2026)
The United States v. Heppner case is the most substantive case to date. Judge Jed Rakoff of the Southern District of New York ordered a criminal defendant to produce 31 documents he had created using Anthropic’s Claude. The defendant, Bradley Heppner, had used Claude to sketch out defense strategies after receiving a federal grand jury subpoena on his own, without direction from his attorney. The court rejected every privilege argument he raised.
The court made three findings of interest particularly for anyone actively engaging in or in the practice of litigation. The findings indicate:
- Claude is not a lawyer. The court found that attorney-client privilege requires a trusting human relationship with a licensed professional. As the court put it, no such relationship “exists, or could exist, between an AI user and a platform such as Claude.”
- The conversation was not confidential. Claude’s terms of service allow Anthropic to collect the user’s inputs and outputs, use them to train its models, and disclose them to third parties. The court found the defendant had no reasonable expectation of confidentiality.
- Acting without attorney direction is critical. Because Heppner used Claude on his own initiative, the court rejected the argument that the outputs were a protected work product prepared at a lawyer’s direction.
Warner v. Gilbarco, Inc. (E.D. Mich. Feb. 10, 2026)
In contrast, a federal court in Michigan achieved a different result on the same day as the bench ruling in Heppner. A pro se employment plaintiff who used ChatGPT to help with her litigation was able to successfully keep the materials protected. The court found that AI tools are “tools, not persons,” and that sharing information with ChatGPT did not amount to disclosure to an adversary, therefore the work-product protection was not waived.
Morgan v. V2X, Inc. (D. Colo. Mar. 30, 2026)
A Colorado federal court extended Warner‘s reasoning and went further. In Morgan v. V2X, Inc. The court protected a pro se plaintiff’s AI-assisted litigation materials as work product and then issued something courts have not previously done: a model protective order barring parties from submitting confidential information to any consumer-grade AI platform, unless the provider was contractually prohibited from training on or disclosing that data. In plain terms, the court acknowledged that mainstream AI tools like ChatGPT and Claude are not safe for confidential case information.
The bottom line across these three cases indicates that the law is unsettled. Heppner offers no protection to a party who uses AI on their own in a criminal matter; however, Warner and Morgan offer stronger protection for pro se civil litigants, but only under specific circumstances and even then, courts are beginning to restrict what information can be fed into these tools at all.
What This Means for Pennsylvania Divorce, Custody, and Support Cases
Pennsylvania’s discovery rules Pa.R.C.P. 4003.1 and 4003.3 are modeled closely on the federal rules. Pennsylvania courts are broadly empowered to require disclosure of any relevant, non-privileged matter. While no Pennsylvania state court has yet ruled directly on AI chat logs in discovery, the federal decisions above apply privilege and work-product principles that Pennsylvania courts also follow, and Pennsylvania practitioners should expect similar arguments to arise throughout the commonwealth.
Family law cases are particularly exposed for several reasons.
Many parties in Pennsylvania family court represent themselves.
Divorce, custody, and support proceedings are among the highest rates of self-represented litigants of any court docket in Pennsylvania. When a parent navigating a custody dispute turns to ChatGPT to understand how Pennsylvania judges evaluate the “best interests of the child” under 23 Pa.C.S. § 5328, or asks Claude what to expect in an equitable distribution hearing, they may be creating a record of their fears, strategy, and intentions. A record that the other side’s attorney could potentially request in discovery.
Even represented parties do their own research.
Pennsylvania family lawyers often hear from clients who spent the night before a court date asking ChatGPT what to expect or asking an AI to explain what their attorney had previously told them. If those conversations included the substance of legal advice or if the client shared facts about the case to get a better answer, they may have undermined the privilege that protects their attorney’s work.
Pennsylvania family law cases involve highly sensitive facts.
Unlike a contract dispute, family law matters require disclosing mental health history, income and assets, parenting habits, and details about a party’s relationship with their children. Pennsylvania financial disclosure obligations in divorce proceedings are already extensive. People tend to share these kinds of details when asking an AI for guidance. And that information, once shared with a third-party platform, may not stay protected.
Discovery requests targeting AI use may be coming.
The court in the Morgan case expressly recognized that opposing parties are beginning to seek AI chat logs in discovery. A motivated opposing party in a Pennsylvania case may request AI interaction logs, prompts, and outputs as a matter of course. Courts evaluating such requests will look at the same factors the federal courts applied: Was the party represented? Did an attorney direct the AI use? What did the platform’s terms of service say about confidentiality?
What You Should Do
None of this means AI tools are off limits for general research. But there are practical steps anyone involved in a Pennsylvania family law matter should take.
- Do not share specific facts about your case. Asking a general question about how Pennsylvania courts divide marital debt is different from describing your finances, what your spouse said in a text, or what your attorney advised. Keep case-specific details out of AI conversations.
- Do not share what your attorney has told you. Disclosing the substance of legal advice to a consumer AI platform is one of the clearest ways to waive attorney-client privilege. If you have questions about what your lawyer said, ask your lawyer.
- Use hypotheticals, not your real facts. You can explore legal concepts by asking about fictional scenarios without using names, dates, or circumstances from your own case.
- Read the platform’s privacy settings. Both ChatGPT and Claude allow users to opt out of some data training. Enterprise-tier accounts offered by these companies come with stronger contractual protections. The Morgan court specifically recognized that consumer-grade AI platforms lack the safeguards that justify placing confidential information in them.
- Talk to your attorney before using AI in connection with your case. The decisions in Warner and Morgan suggest that attorney involvement in AI use can strengthen, though not guarantee, work-product arguments. Document that involvement.
- Assume your chats could end up in court. Until Pennsylvania courts issue clearer guidance, the safest approach is to treat anything you share with a consumer AI platform as potentially discoverable. If you would not want a judge to read it, do not type it into a chatbot.
Frequently Asked Questions About ChatGPT, Privacy, and Pennsylvania Family Law
- Can ChatGPT messages be used in a Pennsylvania divorce case? No Pennsylvania appellate court has yet ruled directly on this issue, but recent federal decisions show that AI chat logs may be discoverable depending on how the tool was used, whether counsel was involved, and whether the information was shared on a consumer platform without meaningful confidentiality protections.
- Should I ask ChatGPT about child custody or support in Pennsylvania? You should be cautious. General legal information may be less risky than sharing your specific facts, but entering details about your custody dispute, finances, parenting history, or your lawyer’s advice may create a record that an opposing party later seeks in discovery.
- Are ChatGPT conversations protected by attorney-client privilege? Not necessarily. Courts analyzing AI tools have emphasized that a chatbot is not your lawyer and that privilege may be lost when confidential information is shared with a third-party platform.
The Bottom Line
Pennsylvania family courts have not yet addressed AI chat logs directly, but the recent federal decisions described above show that this issue is moving quickly. For anyone involved in a Pennsylvania divorce, custody, or support matter, the safest assumption is that chatbot conversations may later be scrutinized if they contain case-specific facts, legal strategy, or confidential communications.
If you have questions about using AI while your family law case is pending, speak with your attorney before entering sensitive information into a consumer chatbot. A short conversation with counsel is far safer than discovering later that your AI prompts or responses may become evidence.
Leech Tishman’s Family Law Practice Group regularly assists families navigating custody disputes, supervised visitation concerns, or petitions to modify custody. Our attorneys are prepared to guide you as courts and practitioners increasingly grapple with the role of artificial intelligence in evidence, communication, and decision-making. For more information about how emerging AI-related issues may affect your family law matter, please contact Gregory F. Suher and Bridget M. Synan, attorneys in Leech Tishman’s Family Law Practice.
Cases Cited
United States v. Heppner, No. 25 CR. 503 (JSR), 2026 WL 436479 (S.D.N.Y. Feb. 17, 2026)
Warner v. Gilbarco, Inc., No. 2:24-CV-12333, 2026 WL 373043 (E.D. Mich. Feb. 10, 2026)
Morgan v. V2X, Inc., No. 25-cv-01991-SKC-MDB, 2026 WL 864223 (D. Colo. Mar. 30, 2026)
Additional Sources
Smith Anderson, “Turning Chats Into Trial Exhibits: Litigation Risks of Generative AI Use” (Apr. 14, 2026) — smithlaw.com/newsroom/publications/turning-chats-into-trial-exhibits
Pa.R.C.P. 4003.1 and 4003.3
23 Pa.C.S. § 5328
Forbes / John Schroyer, “New AI Warning: Don’t Discuss Your Legal Problems With Claude Or ChatGPT” (May 28, 2026)