How to tell whether your AI meeting notetaker actually needs everyone's consent

An AI notetaker joining your call is not automatically legal to run. Federal law sets a low floor -- one participant's consent is enough to record almost anywhere in the country -- but at least a dozen states require every participant's consent, and a federal judge is right now deciding whether a bot's icon in the meeting window even counts as notice. Whether you need to ask before your next call depends on where the people on it actually are, not on what the notetaker's default settings assume.

This isn't a hypothetical compliance exercise. In re Otter.AI Privacy Litigation -- four consolidated federal class actions accusing one of the most widely used AI notetakers of recording and training on calls without every participant's consent -- had its core wiretap and biometric-privacy claims survive Otter's motion to dismiss on August 13, 2026. A second notetaker, Granola, was sued on nearly identical grounds five weeks earlier, over marketing that told users other call participants "won't know it's there." A tool built to make meetings easier to remember is now the subject of the same wiretap statutes written for phone taps.

This applies whether the notetaker is a dedicated app like Otter or Granola, or a feature built into the platform itself -- Google's Gemini in Meet and Microsoft's Copilot in Teams face the same underlying consent question, even though neither has been sued over it yet.

The federal floor, and the states that raise it

Federal law requires only one participant to consent before a call can be recorded -- under 18 U.S.C. § 2511(2)(d), that participant can be you, the host, recording your own meeting. That's the entire national baseline. States are free to require more, and at least a dozen do: everyone on the call has to agree, not just the person who started the recording. When a call crosses state lines -- which most video meetings do -- courts applying the federal Wiretap Act have generally held that the strictest state involved controls the whole call, not just the participants physically located there.

How many states actually require all-party consent depends on which list you read, and that's worth knowing before you trust one. A widely cited count names twelve states, including Oregon -- except Oregon's own all-party rule applies only to in-person conversations; a phone or video call there is one-party consent, the same as the federal floor, which means that list is describing the wrong scenario for the exact case this guide is about. A more conservative legal-industry count lands on eleven: California, Connecticut, Delaware, Florida, Illinois, Maryland, Massachusetts, Montana, New Hampshire, Pennsylvania, and Washington, with Michigan and Nevada sometimes added as a twelfth and thirteenth. The number to trust is your own state's current wiretap statute, not any listicle's tally -- including this one.

Is the bot's icon in the meeting enough?

Otter.ai's own defense, argued before Judge Eumi K. Lee on May 20, 2026, was that a bot's visible presence in the participant list is itself sufficient notice -- that anyone who sees "Otter Notetaker" join the call has been told everything the law requires. It was, by that point, the first time a federal judge had been asked to decide whether a decades-old wiretap statute reaches an AI bot at all.

That distinction -- what a court actually decided versus what a company argued and hasn't yet won -- is the one worth holding onto for your own calls. The ruling that let those claims proceed rested on a narrower theory: that Otter is not simply a tool acting on the host's behalf, but a company that keeps and uses conversations for its own model training, which makes it look more like an eavesdropper than a neutral recorder in the eyes of the law. Whether your notetaker crosses that same line depends on its own data-retention terms, not its icon.

The five-minute check before your next recorded call

None of this requires becoming a wiretap lawyer before your next standup. It requires answering one question honestly -- who's actually on this call, and where -- before you let a notetaker join it silently. The same access-first habit applies to any AI tool with a standing line into your work: it's the same instinct behind connecting an agent to your email and calendar without over-granting. Five steps, about five minutes the first time:

The five checks above assume you're the one running the call. If you're just a participant who spots a notetaker bot and isn't sure it's been cleared with everyone, the response depends on what you actually know:

Whichever situation applies, notice that none of the four branches treat the bot's presence alone as an answer -- because right now, in the one case that's actually tested that argument, it hasn't been one.

None of this is an argument against using an AI notetaker -- transcripts genuinely help, and getting the words right matters too (our companion guide on catching a bad AI transcript covers the accuracy side of the same tools). It's an argument for treating the consent question with the same seriousness you'd give an actual audio recorder in your pocket, because legally, an AI notetaker is exactly that -- with the added twist that nobody has yet decided in court whether its own on-screen presence is enough to ask on your behalf.