Sleep headband makers are training AI on users' brain-wave data

Several companies have not clearly told customers their recordings feed AI models, and corporate data transfers often need no consent.

An investigation by Science with the Pulitzer Center's AI Accountability Network found that several makers of EEG-based sleep and neurotech wearables have built up tens of millions of hours of brain-wave recordings and are using them to develop brain foundation models. Interaxon, maker of the Muse headband, has trained a model on roughly 50,000 hours of its customers' recordings and says identifying details are stripped first, though its privacy policy originally did not mention this use and has since been revised. Emotiv is also training a model partly on customer data and has added an opt-in disclosure. The reporting noted that more than two-thirds of the companies studied allowed corporate transfers of data without consent, a risk if a small firm is acquired, while Colorado and California have extended privacy law to neural data. The article also argues that wearable devices, which use far fewer electrodes than clinical systems, are unlikely to read thoughts despite promising research such as Piramidal's seizure-detection results.