In brief: Companies that direct users' texts, emails, art, online behavior, and other work or private data to advertisers and AI training models have long been under scrutiny. However, a recent report highlights how similar practices in the brain-computer industry have largely flown under the radar. Some experts say consumer brain scanners can't collect much data, but others worry about what might happen if advertisers and AI companies can access users' brain activity.
A recent report from the journal Science outlines the relationship between the electroencephalography (EEG) industry and AI. It found that some companies selling brain-tracking devices aren't entirely forthcoming about where users' brain-signal data ends up.
InteraXon's Muse brain-sensing headband is a leading example. The $475 device tracks and charts users' brainwaves to help improve sleep and focus. But the company didn't inform users that it was using data from some headbands, supposedly anonymized, to train its AI model of the brain until after Science notified them.
Science also examined the privacy policies of dozens of other consumer-facing EEG companies and found that around one-third didn't clearly explain how they handle user brainwave data. Of those companies, 11 had policies that mentioned only how users interact with their websites, and four had no privacy policies at all.
A relatively little-known industry has been exploring EEG devices since long before the AI boom. InteraXon first unveiled what would become Muse in 2010 and spent the next decade promoting it as a project for brain-controlled devices and mental health. However, the AI industry's booming popularity and the investment flowing into it may change how companies use data from brain-tracking wearables.
InteraXon and other companies claim that training AI models on tens of thousands of hours of user data makes their products more helpful. For example, a brain model from AI neurotech startup Piramidal successfully identified nine out of 10 seizures in a pilot study last year.
Some experts note that the data collected by consumer EEG devices is limited in range and usefulness compared to data collected by medical-grade scanners. For instance, Muse headbands use only a handful of electrodes for user comfort, compared to the EEG caps used in labs, which are covered in up to two dozen electrodes. Furthermore, a 2018 study suggests that the research underpinning most EEG products does not meet a high standard.
Meanwhile, other experts are concerned about how much AI models could learn when trained on thousands of people's brainwaves. Although companies can already learn people's thought patterns from what they click, brain data might be the final frontier of privacy.