Plenty of Zoom calls now have an AI notetaker that joins before anyone says hello. The bot pops into the participant list, starts turning everyone's speech into text, and nobody stops to ask if the whole call is fine with it. Now a big group lawsuit in federal court against Otter.ai argues that this design was never OK.
What the lawsuit actually says
Several lawsuits got bundled into one combined case called In re Otter.AI Privacy Litigation. The federal judge handling it is Eumi K. Lee, in the U.S. District Court for the Northern District of California. Otter has asked the court to throw the case out, and that gets heard on May 20, 2026, so there is no ruling yet. The claims are specific. Otter's tool joins meetings on its own, records the audio without a yes from everyone on the call, and feeds those recordings back into its software so the speech-to-text gets better, all without clearly telling the people who do not even have an account.
The case leans on several laws at once. Two federal ones cover it, the Electronic Communications Privacy Act and the Computer Fraud and Abuse Act. State law adds old rules that ban secretly recording private conversations, plus Illinois' Biometric Information Privacy Act, which treats a voiceprint (a digital fingerprint of how your voice sounds) like a body measurement. Under that Illinois law, called BIPA, each slip-up can cost thousands of dollars. Working out who said what in a transcript means building a voiceprint of each speaker. That's the privacy risk most people have not thought about.
Otter's apparent defense is worth noting. According to law firm Fisher Phillips, the company tells customers to "make sure you have the necessary permissions" before using the tool, basically pushing the job of getting everyone's OK onto whoever paid for the account. A strange position, since those account holders had no say over who joined.
tl;dv, Fisher Phillips, Workplace Privacy Report
The all-party consent trap
Federal recording law has a low bar. Just one person on a call can say yes to recording for everyone else. But roughly a dozen U.S. states are stricter, they require every single person in the conversation to agree before it gets recorded. Run one video meeting with people in California, Illinois, and Maryland at once and you may have broken three different sets of rules. Most companies running video calls have not even checked.
Fireflies.ai is facing its own round of lawsuits under that same Illinois voiceprint law. Same pattern. AI notetakers grabbing voiceprints from people who never agreed, across more than one tool, not just Otter.
One in five professionals said they often use AI to write up meeting notes in a 2025 survey. At that scale, this is not some rare edge case. It's a real legal danger for any company running regular video calls.
tl;dv consent breakdown, HR Executive
Where lawyers get nervous
In February 2026, a federal court in the Southern District of New York found that AI-made documents an executive sent to his lawyer were not protected by attorney-client privilege. That's the rule that normally keeps talks with your lawyer secret, so nobody can force you to hand them over in court. It can vanish when a recording bot was sitting in the room.
Courts do not agree on this yet. In Detroit, a federal judge decided that a person handling their own case, with no lawyer, could not be forced to hand over their ChatGPT chats. A federal court in Colorado in March 2026 found some AI messages were shielded as a lawyer's private work, though the specific AI product used was not. Three different courts, three different answers.
The New York City Bar Association put out a 2025 opinion saying lawyers should think hard about whether AI recording tools are even wise to use, and should warn clients about the risks. That's not a glowing review.
New York Times AI Notetakers Legal Risk, ABA Journal Use Humans Not AI Take Meeting, Blank Rome
What employers should do
Nobody fully knows yet. Law firm Littler Mendelson's February 2026 write-up laid out seven trouble spots. Getting consent, voiceprints, accuracy, unfair treatment of certain groups, lawyer-client secrecy, how long data is kept, and confidentiality. That's a lot for one bot to set off.
Unfair treatment is the one most people miss. If a notetaker keeps mishearing accents or speech tied to a protected group, and those messy transcripts then shape performance reviews or hiring, the company could be on the hook for treating people unfairly. New York City, Illinois, and California already make employers tell people and run checks when AI tools touch job decisions.
Banning notetakers outright probably will not work. Employees use personal accounts and the bot shows up whether you want it or not. Littler suggests picking one tool you trust, setting it up carefully, turning off speaker identification where the voiceprint risk is high, and deleting recordings quickly instead of keeping them.
This was always going to happen
Otter.ai is the most visible case right now. The bigger problem is wider than one company. AI meeting bots quietly joined millions of calls over several years before anyone seriously asked whether that was even legal. It's hard to feel sorry for companies that rolled these tools out across the whole org without first checking who had agreed to be recorded. Dropping a recording bot into every meeting and betting that one weak consent rule plus some fine print would cover it was reckless from the start.
