Design AI meeting recaps around recording and privilege risk

The New York City Bar Association issued an ethics opinion cautioning attorneys not to record nonclient conversations with artificial intelligence tools, according to Law360 Legal Industry. The opinion addresses AI meeting recaps as a distinct business-record category and calls for meeting-level controls and framework-level updates, says William Wright at Faegre Drinker.

That moves privilege handling from a back-office security question into the service itself. A recap tool does not merely store a finished document: it can place a conversation into a business-record system whose treatment may matter in discovery and privilege analysis. Law360 Legal Industry reports that the New York City Bar Association opinion cautions against recording nonclient conversations with AI, while Legaltech News (Law.com) says AI agents disrupt the clearer workflow in which an attorney inputs information, software processes it and the attorney reviews the output. An AI-native legal service therefore needs controls at the meeting level, not only a general promise that its systems are secure.

The order of the warning matters. It does not establish that every AI recap is outside privilege, and it does not establish that every legal meeting becomes protected because an AI tool handled it. The legal relationship, the purpose of the communication, the confidentiality arrangements and the handling of data remain the facts that determine the analysis. The opinion instead establishes a narrower point: recording a conversation with AI can create a separate category of business-record and discovery risk that a firm has to govern before the recap is made. Privilege handling is part of the product design because the workflow itself can affect the legal protection at issue.

That is a meaningful constraint on the idea that an AI-native firm can scale by standardizing the underlying software and leaving judgment to a later review. Legaltech News (Law.com) reports that, with a traditional software tool, privilege and confidentiality obligations run through the attorney’s conduct, while AI agents disrupt that model. Legal Futures separately reports that most law firms and in-house legal departments have yet to get on top of AI governance despite widespread use. The combination does not show that AI-native delivery fails. It shows that a viable service must make data handling, recording decisions and accountability visible parts of the client-facing workflow.

The next confirming fact is concrete: a New York City Bar Association opinion, court ruling or discovery dispute that addresses an AI-generated meeting recap and the circumstances of its creation. A firm policy that requires consent or exclusion for nonclient conversations, assigns responsibility for meeting-level settings and explains how recaps are retained would also show the design change this warning demands. We should treat broad claims that AI makes legal communications protected, or that AI necessarily destroys protection, as inadequate until those controls and their consequences are visible in practice.