The Risks of AI Notetaking

While we heard it here first from our very own Dan Tenenbaum (Keep that Bot Out of Your Boardroom), the New York Times’ recent article, “All Those A.I. Note Takers? They’re Making Lawyers Very Nervous,” underscores—and sharpens—a number of risks that companies should be thinking about before inviting AI notetakers into board meetings, strategy sessions, and everyday corporate discussions.

Described as a “ticking time bomb,” AI notetakers create a near-verbatim, searchable, and durable record of every utterance, off-hand remark, inside joke, and speculative comment made in a meeting. In the abstract, that may sound like good governance. In the context of litigation or regulatory inquiry, it can be deeply problematic.

What was once fleeting conversation—never intended to be memorialized—suddenly becomes consequential evidence, stripped of tone, context, and nuance in the cold unforgiving vacuum of legal discovery.

Even small inaccuracies can have outsized consequences. As the Times article highlighted, a simple mistranslation—from “does matter” to “doesn’t matter”—can entirely invert meaning and create years of headaches if that statement surfaces during discovery. AI systems, while improving, still make transcription and interpretation errors, particularly in fast-paced discussions, technical conversations, or where multiple speakers are involved. The combination of AI fallibility and the legal system’s tendency to treat written records as authoritative creates a dangerous mismatch.

More broadly, AI-generated records can amplify normal business rhetoric into something far more problematic. Executives frequently use shorthand, aspirational language, or even exaggerated phrasing in internal discussions—terms like “dominate the market,” “crush competitors,” or “lock up supply.” In the moment, these are often harmless expressions of enthusiasm or strategic ambition. But when captured, preserved, and later reviewed in isolation, they can be recast as evidence of anticompetitive intent or improper motive in an antitrust investigation. Similarly, statements reflecting preliminary thinking, incomplete risk assessments, or evolving strategies may be cited by shareholders as evidence that management downplayed or ignored known risks.

There is also a governance dynamic to consider. The existence of a comprehensive AI-generated transcript may chill candid discussion in the boardroom and among management teams. Directors and officers may become more guarded, less willing to test ideas, or less inclined to voice dissenting views—all of which cuts against effective corporate oversight and deliberation.

Perhaps most concerning, however, is the potential impact on attorney–client privilege. The presence of an AI notetaker—particularly one that involves third-party processing, storage, or access—raises real questions about whether privilege may be waived. If privileged conversations are recorded, transmitted, or stored in a manner accessible to third parties (including the AI provider), companies could face arguments that confidentiality—and therefore privilege—has been compromised. Even if privilege ultimately survives, the fight over whether it applies can be costly, time-consuming, and uncertain.

None of this is to say that note-taking—AI-driven or otherwise—is inherently bad. There is real value in maintaining accurate records of decisions and action items, particularly for operational follow-through and regulatory compliance. But there is a significant difference between intentional, curated recordkeeping and the indiscriminate capture of every word spoken in a meeting. Simply put, in many high-stakes corporate settings, the marginal benefit of AI-generated notes is unlikely to justify the legal risk.  As Dan and the New York Times said, boot the bot!