AI-powered note-taking and transcription apps have quickly moved from niche productivity tools to everyday meeting companions. Yet their rise is now drawing fresh attention to how much data they collect, store, and reuse.
At the center of the discussion is Granola, which is facing a class-action complaint in California. The lawsuit argues that the app can record meetings without clear consent from everyone involved and that its design makes it difficult for other participants to know it is active. The complaint also says users are opted into model training by default, while earlier conversation data may remain available even after settings are changed.
Granola has rejected the allegations, saying it anonymizes training data and does not share conversation data with third parties. Still, the case has become a broader symbol of a fast-growing question in AI: who is responsible for consent when software quietly joins a conversation?
Granola is not alone. Otter.ai has also faced consolidated litigation over claims tied to recording and AI training, while Fireflies.ai has been challenged over speaker-recognition practices. At the same time, several universities, including the University of Washington, have restricted certain AI meeting tools after raising concerns about institutional data security.
The legal debate is especially relevant in states with all-party consent rules, where recording a conversation may require permission from everyone present. As a result, users are being urged to review app settings carefully, disable model training when possible, and clearly inform participants before starting a recording.
Beyond the courtroom, this moment signals a turning point for AI collaboration tools: the next generation of productivity software may need to be built with transparency, consent, and trust at its core.
