Discovering a Conversation with a Machine Friend: AI-Assisted Legal Research as an Unmitigated Litigation Vulnerability

Abdilla, Justin, Discovering a Conversation with a Machine Friend: AI-Assisted Legal Research as an Unmitigated Litigation Vulnerability (February 12, 2026). Available at SSRN: https://ssrn.com/abstract=6227600 or http://dx.doi.org/10.2139/ssrn.6227600

On February 10, 2026, a federal judge ruled that every document a criminal defendant generated using a commercial AI tool was discoverable. The ruling in United States v. Heppner applied existing privilege doctrine to AI-generated legal research and found it protected by neither the attorney-client privilege nor the work product doctrine. The result is architecturally inevitable: because commercial AI platforms route queries through third-party servers, every interaction fails the confidentiality requirement that both doctrines demand. This paper argues that the legal profession’s response — update your policies, train your staff, be careful what you type — is inadequate. It catalogs six attack vectors an adversary can exploit to obtain AI interaction data, demonstrates that analogous research activities already receive legal protection, and proposes a three-part remedial framework: draft amendment language for Federal Rule of Civil Procedure 26(b)(1) establishing qualified protection for AI-assisted research, a technological remediation strategy through local AI deployment, and a practitioner’s objection toolkit with ready-to-file motion language. More than eight hundred million people use generative AI. The discovery apparatus has identified their conversations as a largely unprotected evidentiary resource. This paper offers a framework to close the gap.