A new brain-computer interface (BCI) device has successfully decoded a person’s internal chatter, accurately deciphering up to 74% of imagined sentences. The device requires users to think of a preset password before it can start decoding their thoughts.
Developed by researchers at Stanford University, the BCI system uses artificial-intelligence models to recognize phonemes, or smallest units of speech, in brain activity recordings. When two participants thought of specific phrases, the device correctly interpreted 74% of their sentences, similar to its accuracy for attempted speech in earlier studies.
This breakthrough represents a significant step towards developing BCIs that can accurately decode internal speech, while also providing a way to protect users’ privacy by only decoding when they think of a specific keyword. The system’s potential applications include restoring speech in people with paralysis or limited muscle control, who often struggle with exhausting and uncomfortable speech attempts.
The study was published in Cell on August 14th and has been hailed as “technically impressive and meaningful” by experts in the field.
Source: https://www.nature.com/articles/d41586-025-02589-5