July 6, 2026
A recent article published by Bryan Cave Leighton Paisner LLP on Lexology provides an important warning about the risks of using AI tools in connection with legal matters. Generative AI tools are being used by businesses increasingly, often without the knowledge or direction of their counsel. This carries significant legal risk.
A recent federal court decision, United States v. Heppner, in which a defendant conducted legal research using an AI platform and sent the results to his lawyers without being directed to do so by counsel shows the risk. The court held that the materials were not protected by attorney-client privilege or the work product doctrine. The AI platform’s privacy policy reserved the right to share user data with third parties, eliminating any reasonable expectation of confidentiality. The work product doctrine did not apply because the lawyers did not direct the research.
There are three areas where these risks are particularly significant: internal investigations, AI legal research, and AI-enabled during meetings and conversation recording.
Best Practices for Internal Investigations
Do not use AI to document, summarize, or analyze investigative interviews without first coordinating with legal counsel. Attorney direction is essential to preserve privilege and work product protection.
If you are a non-attorney conducting an AI-assisted interview, confirm that you are doing so at the express direction of legal counsel. You should also ensure that both your use of AI and the fact that you are conducting the interview at the express direction of legal counsel are communicated clearly in any interview warnings given to witnesses.
Be mindful of litigation hold obligations, other preservation requirements, and default retention settings on AI platforms, which may automatically delete inputs and outputs that may be subject to retention obligations.
Always inform your lawyers before using any AI tool to research legal issues, even for background purposes. Lawyers should direct and supervise that research.
Only use enterprise-grade AI platforms on which the vendor contractually agrees not to use your data to train or fine-tune its models and not to disclose your data to third parties. Consumer AI tools typically disclaim confidentiality in their terms of service.
Obtain the express, documented consent of all participants before activating any AI recording or transcription tool during an interview or meeting.
Determine whether the AI tool’s speaker-attribution features may capture or process biometric identifiers, and if so, evaluate compliance obligations under applicable state biometric data privacy laws.
If the AI tool used by your vendor does not operate under a contractual confidentiality commitment, assume that information shared in recorded meetings is not private and limit what is discussed in those sessions accordingly.
To read the full article, click here [1].