AI Meeting Notetakers Offer Efficiency but Raise Privacy Alarms

Artificial intelligence tools designed to take notes and summarize virtual meetings offer the allure of instant recaps and organized to-do lists. Seconds after an hour-long conference concludes, these AI agents can generate key takeaways and action items for participants. However, the methods employed by these popular AI notetakers are prompting caution and concern among professionals.

The core of the unease lies in how these technologies process information. AI notetakers transform all spoken content during meetings into data. This raises alarms because confidential personnel details, corporate strategies, trade secrets, and potentially incriminating remarks could be exposed if this data falls into the wrong hands.

"There are huge risks to the organization on AI notetakers," said Amy Dufrane, CEO of HRCI, a provider of human resources training and certification. "I don’t think companies should use it at all."

An AI notetaker is a software application or device that utilizes artificial intelligence, speech recognition, and large language models to record, transcribe, and summarize conversations. While intended to boost efficiency and engagement, professionals across various sectors are urging vigilance.

Key Concerns About Data and Privacy

A primary concern is the ambiguity surrounding where and for how long the collected data is stored. Privacy advocates worry that companies developing these AI notetakers are creating voiceprints without explicit consent. Voiceprints, a form of biometric data akin to fingerprints but tailored to an individual's vocal characteristics, can be used to access restricted or confidential information, including financial accounts.

Some technology companies have been known to resell data gathered from notetaking tools or to use confidential meeting transcripts and recordings for training their AI models. A significant risk involves attorney-client privilege, as evidenced by a February ruling where a New York federal judge compelled a criminal defendant to share documents created for his lawyers because they had been shared with a third party, Anthropic's Claude.

"People who use AI notetakers, they don’t always know where the data goes," stated Justin Daniels, a corporate attorney at Baker Donelson. "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."

Navigating AI Notetaker Etiquette

When entering a virtual meeting, it is advisable to check for the presence of an AI notetaker. These tools may appear as attendees, often labeled as AI notetakers, or signal their presence via on-screen pop-up messages indicating a recording is in progress. While platforms like Zoom and Google Meet provide clear notifications for recordings, other meeting software may not explicitly indicate when a notetaker is active, according to Thorin Klosowski, senior security and privacy analyst at the Electronic Frontier Foundation.

Participants might also use independent notetaking devices not integrated with the meeting platform, leaving other attendees unaware that a discussion is being recorded and transcribed. "You hope the other person would tell you that they’re doing that," Klosowski said. "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, it is appropriate to inquire. Stating at the outset that a meeting is not authorized for recording can also establish boundaries. Dufrane suggested, "Our company policy is that this meeting cannot be recorded." This approach shifts the responsibility from an individual employee to the company policy.

An alternative strategy involves permitting the AI notetaker for a portion of the meeting, then disabling it for discussions involving sensitive topics. "I won’t start talking about anything substantive until it’s shut off, because I just don’t want to take the risk," Daniels commented.

Understanding Voiceprints and Legal Protections

Many AI notetakers create unique voiceprints for each speaker to distinguish them, often labeling them as "Speaker 1" or "Speaker 2." These voiceprints can be misused for fraudulent purposes if obtained by malicious actors, such as gaining unauthorized access to bank accounts.

Legal frameworks in some states address the creation and storage of voiceprints, granting individuals rights to object to their use in meetings. In Illinois, for instance, voiceprints are classified as biometric identifiers under the state's Biometric Information Privacy Act. This law mandates written notice and informed consent before data collection, along with a defined data retention and destruction policy, according to Chris Pluymers, an associate attorney at The Dillon Law Group. He noted that many companies using these tools lack such systems.

Pluymers emphasized the need for a rigorous approach to AI, data privacy, and biometric identification, stating, "I think getting out ahead of it is crucial."

Under Illinois law, employees can refuse to participate in meetings with AI notetakers if they do not receive assurances regarding data storage, purpose, and deletion timelines. They can also request to see company policies and consent forms.

If an AI notetaker appears unexpectedly, participants can politely request its removal. Pluymers suggested saying, "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." He added, "Just being warm and genuine about it and asking them to respect your wishes."

When using AI notetaking applications, it is crucial to ascertain whether the developers retain recordings, transcripts, or metadata indefinitely or use them for AI model training, advised Danielle Kays, a partner at Fisher Phillips specializing in privacy and employment law.

"If there is some sort of speaker ID or voice recognition, really understand what that is and how it works," Kays urged.

Even if meeting content is deleted, associated metadata can persist with the vendor. This residual data could inadvertently influence AI model behavior or be reproduced. AI notetakers generate text, which is more easily searchable and cheaper to store than audio or video files, making sensitive business information more accessible to unauthorized parties, according to the EFF.