Neurological disorders

Brain Implant Decodes Inner Speech with 74% Accuracy

Published AUG 14 2025

Story

Above: Stanford University campus in Stanford, Calif., on March 28, 2025. Image credit: Justin Sullivan/Getty Images

The Facts

Stanford University researchers developed a brain-computer interface (BCI) that decodes inner speech with up to 74% accuracy. It recorded neural activity from microelectrodes implanted in the motor cortex of four participants with severe paralysis.
Inside Precision Medicine
Nature
Translating the sounds of 125,000 words with AI, they found that attempted and imagined speech activate overlapping motor cortex regions, with inner speech showing similar but weaker neural patterns. Brain recordings revealed that both generate comparable activity when saying or imagining words.
Stanford
The New York Times
Nature
To address privacy concerns, they implemented a password-controlled system that requires users to think of a specific keyword before the device begins translating inner speech. The system recognized the password "chitty chitty bang bang" with over 98% accuracy.
Gizmodo
Nature
Four participants with severe speech impairments were tested, with three finding it easier and more comfortable than attempted speech. It bypassed barriers like breathing control, which typically limits muscle control
Interesting Engineering
Inside Precision Medicine
The BCI sometimes detected unplanned thoughts, like numbers silently counted during visual tasks, showing it could capture spontaneous mental activity beyond intended speech, raising questions about language's role in various thought processes.
Gizmodo
The New York Times
The researchers said that while 74% accuracy is proof of concept for inner speech decoding, it's not yet fluent, adding that they're optimistic that advances in hardware, sensors and algorithms could make communication as natural as regular speech
Interesting Engineering
Stanford

Sources Split

LEFT

CENTER

RIGHT

ESTAB
CRIT

CENTER

PRO
ESTAB

The Spin

Techno-optimist narrative

This study offers unprecedented hope for speech impaired patients. Not only will brain chips be able to decode inner thoughts, but they also show promise for restoring natural speech in people with paralysis. By translating brain activity into near-real-time voice, bypassing physical barriers, these inventions could one day let patients speak fluently again.

Stanford

National Institutes of Health (NIH)

Techno-skeptic narrative

While curing speech paralysis is undeniably a good thing, brain-computer interfaces could also be exploited for surveillance or manipulation. Past abuses, like Cold War mind-control experiments, show how decoding or altering thoughts could threaten mental privacy and autonomy.

Earthling News

Harvard Gazette

Metaculus Prediction

Limited Coverage

This story currently has limited coverage. We will continue to monitor all major outlets and update our reporting as additional information becomes available.