AI Notetakers' Promise Meets Professional Doubt

The allure of artificial intelligence tools that can efficiently take notes and summarize virtual meetings is undeniable. In mere seconds, these AI agents can process hour-long video conferences, providing comprehensive recaps of key points and outlining critical to-do lists for all participants. However, the very methods by which popular AI notetakers achieve these tasks raise significant privacy and security concerns, leading many to avoid their use.

At its core, an AI notetaker is a sophisticated software application or device that leverages artificial intelligence, speech recognition technologies, and large language models to record, transcribe, and summarize spoken conversations. While these tools are designed to enhance productivity by saving time and improving meeting participation, professionals across various fields are issuing warnings due to inherent risks.

A primary concern revolves around the extensive data collection involved. Everything discussed during a meeting—including confidential personnel information, proprietary corporate strategies, valuable trade secrets, and even remarks that could later be deemed incriminating—is converted into data. This digital footprint could potentially fall into the wrong hands, posing substantial risks to organizations. "There are huge risks to the organization on AI notetakers," stated Amy Dufrane, CEO of HRCI, a human resources training and certification provider, who added, "I don’t think companies should use it at all."

The uncertainty surrounding where this collected data is stored and for how long is a major point of contention for privacy advocates. There are concerns that companies behind these AI notetakers might be creating sensitive biometric profiles, specifically "voiceprints," without explicit consent. A voiceprint is a unique biometric identifier, akin to a fingerprint, but tailored to an individual's distinct vocal intonations and characteristics. Such voiceprints could potentially be exploited by malicious actors to access restricted or confidential information, including bank account contents, highlighting the severe implications of unauthorized data collection.

Furthermore, a troubling aspect is that some technology companies are reportedly reselling data harvested from their notetaking tools or utilizing confidential meeting transcripts and recordings to train their proprietary AI models. This practice not only breaches trust but also raises ethical questions about data ownership and usage. The legal ramifications are also significant; conversations protected by attorney-client privilege could lose their confidential status if shared with a third-party AI notetaker. An instance of this occurred in February when a New York federal judge ordered a criminal defendant to provide documents shared with Anthropic’s Claude, because they were no longer deemed privileged after third-party disclosure. Justin Daniels, an Atlanta-based corporate attorney at Baker Donelson, emphasized this point, stating, "People who use AI notetakers, they don’t always know where the data goes. And in my context, if the data goes anywhere else and they’re not aware of it, that attorney-client-privileged conversation may not be attorney-client-privileged anymore."

To mitigate these risks, individuals and organizations must proactively manage the presence of AI notetakers in virtual meetings. The first step involves habitually checking for bots upon joining a meeting. These AI notetakers often appear as a named attendee, such as "AI Notetaker," or a pop-up message may indicate that the meeting is being recorded. While platforms like Zoom and Google Meet typically notify users when recording is in progress, some meeting software may not clearly disclose the presence of a notetaker. Moreover, participants might use personal recording devices separate from the meeting platform, making it crucial to establish communication.

Thorin Klosowski, senior security and privacy analyst at the Electronic Frontier Foundation, recommends that "Asking everyone for consent before doing a sensitive meeting would be the most polite approach to take." If unsure about an AI notetaker's presence, participants should directly ask. Alternatively, one can explicitly state at the outset that a meeting is not authorized for recording. Amy Dufrane suggests a polite way to set this boundary: "Our company policy is that this meeting cannot be recorded," which places the onus on company policy rather than an individual. Another strategy is to allow the notetaker for less sensitive portions of a gathering but insist it be turned off for delicate discussions. Daniels noted, "I won’t start talking about anything substantive until it’s shut off, because I just don’t want to take the risk."

Asserting privacy rights, particularly concerning voiceprints, is also crucial. Chris Pluymers, an associate attorney at The Dillon Law Group, explains that many AI notetakers determine unique acoustic signatures for each speaker to differentiate them, often labeling them as "Speaker 1" or "Speaker 2." The misuse of these vocal signatures, if acquired by bad actors, could lead to unauthorized access to files, fraud, or account takeovers. State laws are beginning to address this; for example, Illinois’s Biometric Information Privacy Act considers voiceprints as biometric identifiers and mandates written notice and informed consent before collection, along with documented data retention and destruction policies. Pluymers points out that most companies using these tools currently lack such systems, underscoring the need for a more proactive and rigorous approach to data privacy in the age of AI.

Under Illinois law, employees have the right to refuse to attend a meeting with an AI notetaker until they receive assurances regarding data storage, purpose, and deletion. They can also request a policy and a written consent form. If an AI notetaker appears unexpectedly, a participant can politely state, "I prefer we keep this meeting without AI recording or transcript tools and I’d be happy to take my own notes and share a recap if that’s helpful," as suggested by Pluymers, emphasizing a warm and genuine request for respect of their wishes.

Finally, understanding where data goes is paramount. Danielle Kays, a partner at Fisher Phillips specializing in privacy and employment law, advises individuals using AI notetaking apps to investigate whether the creators retain recordings, transcripts, or metadata indefinitely, or if they use them to train AI models. It is essential to "really understand what that is and how it works" if speaker ID or voice recognition is involved. Even after content deletion, metadata might persist with the vendor, potentially influencing the AI model's behavior and, in some cases, leading to the memorization or reproduction of sensitive business information. The Electronic Frontier Foundation also notes that AI notetakers generate text, which is significantly easier and cheaper for outsiders to search and store compared to video or audio files, further increasing the risk exposure. "Storing a bunch of video isn’t easy, it’s costly and hard to look through, but text is much easier to search and cheaper to store," said Klosowski. Vigilance and informed decision-making are critical when engaging with these advanced, yet potentially perilous, technologies.