AI Use Is Expanding—And Courts Are Treating Client-Generated AI Prompts as Discoverable Evidence
As generative AI tools become embedded in everyday business operations, courts across the United States are increasingly being asked to determine whether AI-generated content—and the prompts used to produce it—must be disclosed in litigation. An emerging trend in federal courts is becoming clear: AI use creates records, and those records may be discoverable.
For Pennsylvania litigants, this development carries significant implications. AI tools are no longer experimental novelties; they are mainstream business utilities. Like emails, text messages, and internal memoranda, AI prompts and outputs may become evidence in litigation.
Courts Are Beginning to Define the Discoverability of AI Prompts
One of the most instructive decisions comes from the U.S. District Court for the Southern District of New York in In re: New York City Bus Tour Antitrust Litigation, No. 14-CV-9180 (S.D.N.Y. Apr. 18, 2024). There, the court held that AI prompts created by a client, without attorney involvement, were discoverable. The party resisting production argued that the prompts were protected attorney work product. The court rejected that argument, explaining:
"The mere use of an advanced tool does not transform a client's independent work into attorney work product."
Because the prompts were generated by the client alone, the court treated them like any other client-created document: discoverable if relevant and not otherwise protected by privilege.
This reasoning is consistent with other emerging federal decisions. For example:
In United States v. Morgan, No. 2:23-cr-157 (W.D. Pa. Feb. 2024), the court addressed whether AI-generated summaries used by investigators were discoverable. The court required production, emphasizing that AI-generated materials are generally evaluated under the same discovery principles as other government-created documents.
In In re: Meta Pixel Healthcare Litigation, No. 22-cv-03580 (N.D. Cal. 2024), the court considered whether AI-assisted data-processing tools generated discoverable metadata. Although the decision focused primarily on privacy issues, the court acknowledged that AI-generated logs and outputs may constitute discoverable electronically stored information (ESI).
Taken together, these decisions suggest that courts are increasingly treating AI-generated materials within the framework of existing discovery rules rather than as a separate category of evidence.
Why This Matters for Pennsylvania Litigants
Although Pennsylvania appellate courts have not yet issued extensive guidance regarding AI-related discovery, litigants should anticipate that similar principles may apply. As AI becomes a routine part of business operations, several litigation risks become increasingly significant:
Privilege concerns. Clients who use AI tools to analyze legal issues before consulting counsel may inadvertently create discoverable records that reveal strategy or legal misunderstandings.
Inadvertent disclosure. Many AI platforms retain prompts and outputs, potentially resulting in sensitive information being stored by third-party vendors.
Accuracy and hallucination risks. AI-generated content may contain factual inaccuracies or fabricated information that opposing counsel can exploit.
Metadata exposure. AI prompts may reveal timing, user identity, or internal workflows that opposing parties may use to reconstruct decision-making or challenge credibility.
In short, AI use can unintentionally create a detailed record of a party's internal thinking, business processes, or legal analyses—records that may later become the subject of discovery requests.
Practical Steps for Pennsylvania Law Firms and Clients
To navigate this evolving landscape, Pennsylvania litigants should consider adopting proactive measures:
Implement an AI-use policy that distinguishes between attorney-directed and client-initiated AI use.
Educate clients about the potential discoverability of AI prompts and the importance of consulting counsel before using AI for legal or quasi-legal tasks.
Document attorney involvement when AI is used as part of legal work to strengthen claims of work-product protection where appropriate.
Review AI platform privacy policies and terms of service to understand whether client data is stored, reused, or shared.
Preserve AI-related materials—including prompts, outputs, and relevant system settings—when litigation is reasonably anticipated.
These practices can help preserve privilege, reduce inadvertent disclosures, and ensure that AI is used responsibly within the bounds of existing discovery obligations.
The Bottom Line
Federal courts are increasingly treating AI prompts and outputs as potentially discoverable materials, particularly when they are created without attorney involvement. The decision in In re: New York City Bus Tour Antitrust Litigation is among the earliest and most prominent examples of this developing body of law. As AI becomes a routine part of business operations, litigants should assume that AI-generated content may be subject to the same discovery obligations that apply to other forms of electronically stored information.
Questions About AI and Litigation?
As businesses increasingly integrate AI into their operations, understanding the legal implications of AI-generated content is becoming essential. If you have questions about developing AI policies, preserving privilege, or managing discovery obligations, the attorneys at Hamburg, Rubin, Mullin, Maxwell & Lupin are available to help.
This Legal Alert is provided for informational purposes only and does not constitute legal advice. Every legal matter is unique and should be evaluated based on its specific facts and circumstances.