The Meeting Room You Specified Might Now Be a Defendant

A conference room doesn't know it's being sued. It just has a mic array in the ceiling, a scheduling panel by the door, and a notetaker bot that joins every call because of the way it was configured, months or years before anyone sitting in the room today had a say in its setup.

Here's what that looks like when it goes wrong. A woman named Chaka Theus got on a Zoom call with a medical professional. Her lawsuit says the conversation touched on what her complaint calls "deeply personal and private medical information," the kind of thing you say to a doctor, not to a room full of strangers.

Otter.ai's notetaker was on that call too, quietly transcribing. According to her complaint, the company didn't just hand her a transcript and move on: it kept the recording and folded it into the data it uses to train the speech-recognition and machine-learning systems it sells.

Theus's case isn't an outlier. The developers of three  AI notetakers, Otter, Granola and Fireflies, now face  lawsuits built on three different legal theories, while a fourth, Read AI, has been restricted by two universities without a single court weighing in. What ties these lawsuits together isn't the apps themselves  it's who let them into the room in the first place, usually not the person running the meeting. As courts start treating these vendors as more than passive tools that happened to be invited in, the open question is how far that reasoning reaches, toward whoever specced the room to begin with.

Theus’s allegation that Otter kept and reused what it heard just cleared its first real legal test, along with the claims of six other plaintiffs alongside hers. On 13 August 2026, a federal judge in California ruled that Theus and most of her co-plaintiffs can continue  their case against Otter, in a decision that treats the company’s app less like a tool the meeting host was using and more like an eavesdropper that  happened to be in the room.

Otter's defense, in essence, was that it was just doing what it was told. Someone invited it onto the call, someone hit record, and whatever consent that person gave should cover Otter too. Judge Eumi K. Lee of the U.S. District Court for the Northern District of California rejected that argument, at least for now, in In re Otter.AI Privacy Litigation. Because the complaint plausibly alleges Otter "independently collects, retains, and uses communications for its own commercial purposes," that's enough, at this early stage, to treat the company as a potential third party to the call rather than a neutral extension of whoever pressed the button.

Otter isn't alone in this fight, and the other two cases work on completely different mechanisms, which is part of why this looks less like a fluke and more like a pattern.

Granola doesn't have a visible meeting bot at all. That's the whole point of its pitch, and the company's own marketing says so directly: "Other people in the room won't know it's there." 

A lawsuit filed in July 2026 alleges that's exactly what happened to one plaintiff. She never used Granola. She never agreed to its terms of service. She was on a call with someone who had, and the app allegedly recorded her anyway, silently, running off another person's subscription. Her complaint quotes Granola's own marketing language back at the company.

Fireflies raises a different problem entirely: not what it records, but what it turns a person’s voice into. Its Speaker Recognition feature allegedly analyzes a recording and generates a voiceprint, a biometric fingerprint of your voice, that is treated by the law much like an actual fingerprint or a retina scan. 

Illinois has one of the strictest biometric privacy laws in the country, and it requires written notice and consent before anyone collects information such as a voiceprint, not just from the account holder, but from everyone on the call. 

Plaintiff Katelin Cruz filed the first case over this in December 2025, alleging Fireflies built a voiceprint from a meeting she attended in November 2025 without ever telling her. She dropped that case in March 2026. The day before she did, a second plaintiff, Ethan Fricker, filed the same allegation in a different Illinois federal court, and that case is still active.

None of these three cases has resulted in a finding of liability. Nobody has been found liable for anything. But three different products, using three different mechanisms, are now being tested against the same basic question: Is it enough that a piece of software was welcomed into the room?

Not every notetaker is being sued. Read AI has never appeared in these lawsuits. Two universities restricted it anyway, and the reason they gave says a lot about why people find this technology unsettling even without a court forcing the issue.

Chapman University banned it outright in 2025, and a separate post on the school's own IT blog, titled "Being stalked by Read.ai?", gets at why in plainer language than any court filing could. People had granted the app permission to join their meetings at some point, forgotten they'd done it, and then found it showing up anyway. Worse, removing the Zoom app from your account doesn't stop it. You have to actively delete your Read AI account to make it go away. 

The University of Washington took a narrower step, blocking Read AI's access to Zoom and moving to do the same for Teams, citing the tool's ability to "join, transcribe, and summarize your online meetings even when the users are not in attendance, and without the awareness or consent of other attendees."

Neither university waited for a judge to weigh in. They looked at what the software could do and decided that was reason enough.

Every story written about this wave of lawsuits, and there have been a lot of them, treats the situation as a series of questions for whoever hosts the meeting: did you get consent, did you disclose the bot, did you read the terms of service. Almost none of them ask who built the room the meeting happens in.

A meeting host doesn't spec the hardware. Someone else, an AV integrator, an AV/IT convergence specialist, chose the microphones, configured the Teams Room or Zoom Room platform, and in a lot of deployments, decided whether a notetaker gets bundled in as a default rather than something a client has to ask for. 

The question a court just spent dozens of pages wrestling with- whether a company that captures and reuses a conversation counts as more than a passive tool– is a question about defaults and configuration as much as it's a question about who clicked record. Nobody has asked, on the record, whether that reasoning reaches back to whoever set those defaults in the first place.

None of this waited for a judge to rule. Months before Judge Lee's order, Microsoft, Zoom and Google all quietly rewrote how their own platforms handle outside bots, moving the decision out of the meeting host's hands entirely. Microsoft Teams now flags any third-party bot as "Unverified" in the meeting lobby and blocks it without the organizer's explicit approval by default, with an admin-level policy that can shut outside bots out of an entire organization at once. Zoom now requires a host or co-host to start its own AI Companion before it can join, and gives administrators a switch to disable it in any third-party meeting. Google Meet's note-taking feature needs that same host approval, and Workspace admins can restrict it by department or team.

Microsoft didn't dress up the reasoning. In a notice sent to IT administrators two months before Otter's hearing, the company warned that bots "may access meetings without the knowledge or consent of the meeting organizer or the hosting tenant, which can create data security, privacy, and compliance risks," language that lands remarkably close to what the court would later find. Metrigy analyst Irwin Lazar put the commercial logic more bluntly: native, platform-built AI "eliminates the need for...third-party meeting recording and transcribing apps like Otter and Fireflies" in the first place.

None of these settings arrived because a court ordered them. They arrived because three of the largest platforms on earth read the same allegations everyone else was reading and decided the safer place to answer the liability question was an admin console, not a courtroom. That console- tenant-wide bot policies, organization-level restrictions, default-on approval requirements- is exactly the layer where an AV integrator configuring a Teams Room or Zoom Room for a client already works. Nobody's called it part of the job yet. The platforms just started acting as if they might be.

For now, the defendants are software companies, not the people who install their products. If more cases like this keep coming, that might not stay true for long. Someone chose the tools, set the defaults, approved the integrations, and handed the finished room over to a client who never had to think about any of it. 

At some point, the people who spec and install those rooms may have to answer a version of the same question the courts are now asking the notetakers: what exactly did you put in this room, and who did you tell?