Long-term independent use of an intracortical brain–computer interface for speech and cursor control

The 60-second view

What problem?

Speech BCIs had been accurate in sessions that researchers ran. Whether one could be used alone at home, day after day and for years, had not been shown.

What did they do?

Followed one man with ALS and severe dysarthria as he used a speech and cursor BCI at home on his own, almost every day, and recorded how much he said, how often it was decoded correctly and what he used it for.

What changed?

The evidence now covers a working life as well as a test session: 183,060 sentences in nearly two years, a full-time job kept, and his own computer operated with decoded speech as the keyboard and a decoded cursor as the mouse.

Why does it matter?

Independent operation at home and reliability over years were the two things earlier speech results had left open. They are now shown in one person; a cohort and a fully implanted device are still to come.

Method

Participants

One man with ALS, paralysis and severe dysarthria

Interface

Intracortical microelectrode arrays, within the BrainGate2 study

Signal

Intracortical neural activity during attempted speech and movement

Task

Everyday communication and computer use at home, with formal tests on prompted words

Decoder

A brain-to-text speech decoder and a cursor decoder, used as keyboard and mouse input

Results

>3,800 hours

Use at home, no researcher present

Over nearly two years

183,060 sentences

Communicated

1,960,163 words

56 words/min

Average rate

>99%

Word accuracy, 125,000-word vocabulary

Formal tests on prompted words

92%

Sentences rated at least mostly correct

The participant's own rating

Strength of evidencePeer reviewed, one participant, with nearly two years and more than 3,800 hours of use at home and no researcher present.

Figures are quoted from the published abstract and were checked on 11 Oct 2026.

Limitations

• A single participant; how far the result carries to other people and other causes of speech loss is not known.
• An investigational system inside a feasibility study, not a fully implanted device.
• The 92% figure is the participant's own judgement of his sentences, not an independent measure.

What this could enable

Judging a speech BCI the way assistive technology is judged: by hours of unsupervised use, and by whether the person keeps using it.