A privacy lawsuit brought against audio transcription service Otter.ai accuses its AI notetaker of feeding the contents of Google Meet, Zoom, and Microsoft Teams meetings into its training data without proper notification and permission.
The suit centers on the Otter Notetaker feature, an optional service that will automatically transcribe voices in video meetings to text. While the Otter privacy policy makes clear that the service may indeed train its AI on the voices of users while they are in meetings, the privacy lawsuit notes that guests without Otter accounts that can be invited to these meetings have not been similarly notified or opted in.
Privacy lawsuit takes Otter to task over guest recording
Both the Otter privacy policy and the “Privacy & Security FAQ” listed at the company website do include segments informing users that their voices can be used for training data, including while participating in meetings. However, these meetings can include guests who do not have an Otter account and would not have had occasion to encounter these warnings prior to being included. It thus asserts that the AI notetaker is in violation of Electronic Communications Privacy Act, Computer Fraud and Abuse Act, and some local California laws regarding permissions and knowledge required for voice recording.
Otter’s AI notetaker service predates the ChatGPT-driven LLM boom of recent years; the company has offered a voice-to-text transcription app for nearly 10 years, racking up a large user base given its reliability and low hardware requirements on phones. The video meeting transcription service was first added in 2020 with Zoom compatibility, but was only made available to those who pay for an account. While it was not made available to free-tier users, a perk that has been added for paying users is the ability to add some amount of guests so as to allow for all participants in a video conference meeting to have their voice utterances translated to text.
The privacy lawsuit was filed by a single plaintiff in the Northern District of California but is seeking to become a class action and has already contacted over 100 potential participants. It asserts that Otter has a responsibility to ensure that its AI notetaker service is in compliance with the law and that this cannot be shifted to its users. California’s “two-party consent” law requires all participants in a recorded conversation to be notified and to provide their permission for the recording, with fines of up to $5,000 per impacted party allowed in civil cases.
AI notetaker sees major growth even as concerns have piled up
The AI notetaker service has seen very substantial growth recently, passing $100 million in annual recurring revenue for the first time in March of this year and seeing its global user base jump to about 25 million (nearly twice what it had just two years ago). The service is also thought to have transcribed over one billion meetings as of this year. But this success has certainly not come without growing pains, and an increasing collection of user complaints.
Users have noted that the service has some rather pushy default features that can opt users into all meetings by default, to include activating the AI notetaker and sending out invitations when not intended to. This can mean that if even one user at a company has Otter active, it may insert itself into various meetings through them and transcribe when other users are not aware that it is active (and are potentially engaging in private conversations not meant to be transcribed or disseminated). These settings can be manually disabled, but are generally on by default and may well go overlooked by less technical users.
The privacy lawsuit plaintiff, Justin Brewer of San Jacinto, CA, notes that the issue is not just technical violation of “two-party consent” laws but also that the AI notetaker is capable of massive privacy invasion of this sort. Users who remain in a meeting after other parties leave may well speak frankly and off-record without being aware that a formerly-present Otter user’s transcription is still running, and may then be further surprised when their off-record comment transcription is automatically emailed to all participants.
In addition to causing such problems in the business world and courting a privacy lawsuit, concerns have been raised about other applications of the AI notetaker. A recent Politico story documented a journalist’s use of it while interviewing an Uyghur human rights activist based in China, and only realizing after the fact that Otter’s privacy policy indicates that it shares data with third parties. Should the Chinese government be one of those data buyers, it could have put the activist at risk; Otter responded to the story with a statement that it does not share data with foreign governments or their law enforcement agencies.
