Your AI Notetaker Is Listening to More Than the Meeting

Somewhere in the last eighteen months, most meetings picked up a new attendee. It does not talk, it does not blink, and until recently almost nobody stopped to ask if it should be there. An AI notetaker joins the call, transcribes every word, and quietly builds a permanent record of a conversation that used to disappear the moment the call ended. A class action lawsuit against Otter.ai, new obligations under the EU AI Act that took effect this month for AI used to monitor workers, and a wave of coverage over AI notetaker privacy concerns in Bloomberg and The Washington Post all point to the same thing: the tool leaders adopted for convenience has become a trust problem nobody budgeted time to solve.

What is actually happening with AI notetakers right now?

The lawsuit against Otter.ai alleges the platform recorded conversations without the consent of everyone on the call and used those recordings to train its models without adequate disclosure. Employment attorneys are calling it one of the hottest issues on their desks this year, not because the technology is new, but because adoption outran the norms around it. A tool that started as a convenience for one distracted note-taker is now capturing performance conversations, salary discussions, and offhand comments that were never meant to outlive the meeting.

The EU AI Act adds a second layer. Beginning this month, AI systems used to monitor workers can be classified as high-risk, which brings real compliance obligations for any organization operating across borders. Meanwhile in the United States, consent law is a patchwork: some states require everyone on the call to agree to being recorded, others only require one person to know. Most AI notetakers default to recording the moment they join, and most employees have no idea which law applies to the room they are sitting in.

Is it legal to record a meeting with an AI notetaker?

It depends entirely on where the people on the call are sitting, and that is exactly the problem. Two-party consent states require every participant to agree before a conversation can be recorded. One-party consent states only require one person, often the host, to know. A leadership team spread across a few states or a few countries can have a legally compliant recording in one seat and a legal violation in the next, inside the same call. Almost no organization has actually mapped this out. Most just accepted the default settings on the tool their team started using because it was free and it saved someone twenty minutes of typing.

Why does this erode trust even when it is technically legal?

Because legal and safe are not the same thing, and people can feel the difference before they can name it. A meeting where everything said becomes permanent, searchable, and shareable is a different meeting than one where people could think out loud, walk back a bad idea, or admit they do not have an answer yet. Teams have started asking to be told before a recording starts, and that request is not paranoia. It is a reasonable reaction to watching honest conversation get harder the moment someone realizes every word might resurface in a dispute, a performance review, or a training dataset months later.

This is where AI in the workplace keeps making the same mistake. The tool exists to serve a human point: give people back the time they spent typing notes so they can actually listen. Instead, in most rollouts, nobody asked what the tool would cost in candor. A notetaker that makes people perform for the transcript instead of speaking freely has stopped serving the room. It has started managing it.

What should leaders actually do about AI notetaker privacy concerns?

Start by making the default visible instead of automatic. Every meeting where an AI notetaker is present should say so out loud, at the start, before anyone assumes the room is off the record. That single habit fixes most of the consent problem and most of the trust problem at the same time, because people stop guessing whether they are being recorded and start deciding, in real time, what they actually want to say.

Second, write an actual policy, not a settings default inherited from whichever tool someone signed up for first. Decide who can turn a notetaker on, who gets notified, how long the transcript lives, and who can access it later. A policy that fits on one page beats a compliance document nobody reads, as long as it answers those four questions plainly.

Third, build in a real off-the-record option. Some conversations, a hard performance conversation, a team airing a real concern, a leader admitting uncertainty out loud, need a room where nothing is being captured. If every meeting defaults to recorded, you have quietly eliminated the only space where people used to say what they actually think. Protect that space on purpose, or it disappears by accident.

Fourth, train the people using these tools, not just the people buying them. Most AI notetaker rollouts happen through IT or a single enthusiastic manager, with no broader conversation about consent, storage, or what happens if a transcript ends up somewhere it should not. A team that understands what the tool does, and what it does not get to do, uses it with judgment instead of habit.

Where this actually leaves your team

The AI notetaker is not the problem. The silence around it is. A leader who adopts a tool that captures every word without deciding, out loud and on purpose, what that means for trust in the room has made a choice by default that deserved to be made on purpose. The invisible cost of an AI notetaker is not a line item. It is every conversation that got a little more careful, a little more guarded, the day people realized they were always being recorded and nobody had told them the rules.

That is exactly the kind of gap The Spotlight Machine’s training sessions are built to close. Aziz Aghayev works with leadership and corporate teams to build the judgment an AI rollout skips when it adopts a tool by default and hopes trust survives the fallout. If your meetings picked up a new attendee and nobody wrote the rules for it yet, that conversation is worth having before the next one starts.