Muse Maker Used 50,000 Hours of Users’ Brainwave Data to Train AI Without Explicit Disclosure
Regulatory & Policy · 4 sources · Updated 6d ago · First seen Nov 9, 2026

Interaxon, the Canadian maker of Muse sleep and meditation headbands, used approximately 50,000 hours of users’ electroencephalography (EEG) recordings to train a brain foundation model without explicitly informing users, a year-long Science investigation found. The company removed names and email addresses before training and later updated its privacy policy to disclose machine-learning use of de-identified data. The investigation also examined Emotiv, which subsequently updated its disclosures to specify opt-in consent for training on de-identified neural data.
Research cited in the reporting indicates that removing direct identifiers from EEG recordings may not prevent re-identification or sensitive inferences, although it does not establish what can be inferred from Muse recordings specifically. A review of 57 brain-data companies found substantial privacy-disclosure gaps and provisions that could permit data sales or transfers to new owners without user consent. Separately, an engineer reportedly discovered hardcoded credentials in a smart sleep mask’s Android app that enabled access to other users’ brainwave recordings. Colorado and California have extended consumer privacy protections to neural data; users and organizations assessing these devices should verify AI-training consent, opt-out and deletion options, and ownership-transfer terms.


