A man with ALS has used his speech brain implant at home for nearly two years, mostly on his own

By Inside BCI · June 16, 2026

Casey Harrell, who has ALS and is paralysed, can speak to his seven-year-old daughter, read her bedtime stories, surf the web, send emails, and work full-time as an environmental activist. He does all of this through an intracortical brain-computer interface implanted in his speech motor cortex in July 2023. In the first 22.6 months after surgery, Harrell used the system at home for more than 3,800 hours without any researchers present.

His team at the University of California, Davis published the result in Nature Medicine on 15 June 2026. The corresponding author is David Brandman, an associate professor of neurological surgery at UC Davis. Neuroengineer Sergey Stavisky, also at UC Davis and a member of the team, is quoted calling Harrell “the first power user of a speech BCI.” MIT Technology Review covered the paper the same day. The trial is registered as NCT00912041, part of the long-running BrainGate2 programme.

How the system works

The device records from the speech motor cortex, the brain region that controls the muscles used to speak. Four arrays of 64 microelectrodes each (256 electrodes in total) were implanted during a five-hour operation in July 2023. The arrays connect through two pedestals on the exterior of Harrell’s skull. Software developed by the UC Davis team translates the recorded neural activity first into phonemes (the 39 sounds that make up American English) and then into words. The team got the speech decoder working on the first day Harrell tried it. By that afternoon, he was using a 50-word vocabulary and 99.6 per cent of the decoded words were what he had intended to say.

How accuracy and vocabulary scaled

The vocabulary was later expanded to 125,000 words at 97.5 per cent accuracy. During his at-home use over the 22 months covered by the new paper, Harrell himself rated 92 per cent of the sentences he produced as accurate or mostly correct. Stavisky told MIT Technology Review that the system’s current controlled-test accuracy is now 99 per cent. The team has also added a cursor decoder, which lets Harrell use a personal computer for messaging, email, web browsing, and his job.

Independence

In 2023, members of the research team had to visit Harrell’s home to physically connect him to the device on days he wanted to use it. Two years later, Harrell’s care partner handles the connection and disconnection. Harrell “wakes up, gets plugged in, and just gets going,” Stavisky told MIT Technology Review. The system now includes a privacy mode that automatically deletes any decoded text while active, and a profanity filter Harrell uses when speaking to his daughter.

Why this matters beyond the single case

Mariska Vansteesel, a BCI researcher at Utrecht Medical Center who was not involved in the UC Davis trial, told MIT Technology Review that “for these technologies to be relevant for patients, we really need to test them in settings in which they will eventually be used … to demonstrate that it has value, that it’s usable, and that it functions well without the constant involvement of a research team.” Until the Harrell paper, no speech-BCI cohort had documented sustained at-home use at this scale. The closest comparable case, also reported by Vansteesel’s group, was a woman with ALS who used a fully implanted system for roughly seven years before it stopped working, apparently due to brain degeneration.

What’s still uncertain

Harrell’s case is one patient. Vansteesel cautioned in her comments to MIT Technology Review that durability and accuracy in a single ALS recipient may not transfer to others, particularly as the disease progresses and brain tissue degenerates. Jane Huggins, who develops non-invasive BCIs at the University of Michigan and was not involved in the trial, told MIT Technology Review that “a consistent aversion to hospital stays among people with progressive conditions like ALS” is itself a deployment barrier for invasive devices. The next several years of BrainGate2 enrolment data and the wider clinical-stage cohort (Neuralink, Synchron, Precision Neuroscience, Paradromics) will show whether the durability and independence Harrell has experienced extend to other recipients.

What comes next

The UC Davis team is developing a “brain-to-voice” system that would decode neural activity into a synthesised voice with natural cadence, inflection, and intonation. Harrell’s current system speaks the decoded words through computer-generated speech. The next-generation aim is for his decoded output to carry the emotional register of voice, including the difference between sounding happy, angry, or sarcastic. The Nature Medicine paper does not give a timeline.

Twenty-two months in, Harrell’s case is the longest documented at-home use of a speech BCI on record, with the most autonomy from research team support so far reported. The result is one patient. It is also the first clean test of whether a speech BCI can sustain real life at home after the research team goes away. The result, for Harrell, is that it can.

Sources

research brain-computer-interface BCI intracortical-BCI speech-BCI ALS Casey-Harrell UC-Davis-Neuroprosthetics-Lab David-Brandman Sergey-Stavisky Nicholas-Card Nature-Medicine BrainGate2