For most of history, people living with amyotrophic lateral sclerosis (ALS) and related neurological disorders have faced a grim outlook for their ability to communicate. ALS, which attacks the nerve cells responsible for carrying signals from the brain to the muscles necessary for speech, can make speech slurred, difficult to understand, and eventually impossible for some patients. Over the past several decades, modern assistive communication technology has made major strides toward enabling people with ALS to communicate, but the technology has fallen short of reproducing the ease and spontaneity of natural communication. An experimental AI-powered tool is changing that.
A new AI-powered brain-computer interface (BCI) system developed at UC Davis in collaboration with researchers at Brown University and Mass General Brigham Neuroscience Institute appears to represent a major breakthrough in addressing two of the “key challenges” to the real-world adoption of this emerging BCI technology: independent at-home use and reliable long-term performance. A BCI is a system that measures brain activity and translates it into commands for an external device without relying on muscle movements normally required to operate a computer.
What does this breakthrough mean for the everyday lives of ALS patients experiencing severe paralysis? A significant leap toward independence and more natural communication.
The latest BCI breakthrough enabled study participant Casey Harrell to communicate substantially more independently than before, with the study documenting more than 3,800 hours of at-home use. And while a caregiver helped set up the system, at-home use was not dependent on the researcher supervision his earlier BCI use required. This allows for sustained and far more natural communication with friends, family, and colleagues. The BCI also enabled him to control a computer cursor, allowing him to navigate the internet, send messages, and perform work-related tasks. Together, these capabilities allow Harrell to participate considerably more independently in family life and work.
Using a 125,000-word vocabulary, the BCI demonstrated 99.2 percent word accuracy in a controlled word-copying task, according to a new June 2026 study. And the pace of this innovation is difficult to overstate. Two earlier studies, one published in August 2024 with the same participant and the other in August 2023 involving a different participant, reported 97.5 percent word accuracy and a 23.8 percent word error rate, respectively, under different testing conditions.
In the June 2026 study, Harrell communicated at an average of 56 words per minute (WPM) using a BCI, compared with roughly 6.3 WPM using head-controlled typing, as documented in an earlier study. For context, conversational English is roughly 160 WPM. And while this remains substantially slower than conversational English, it represents a meaningful jump over less than three years.
From a technical standpoint, this BCI works by using implanted electrodes—tiny sensors that record electrical activity in the brain—to send those recordings to AI decoding software capable of translating those signals into text. From there, those words can be displayed or played aloud. This is a break from traditional assistive systems that rely on things like movements of the eyes, head, and other parts of the body to select words and letters.
Recent AI-powered advancements in BCI technology do not stop at helping users communicate with greater speed, accuracy, and independence. Another experimental approach is pushing the boundaries of restoring natural-feeling communication, translating brain activity not only into words, but tone, emphasis, and even song. And it’s doing so in real time.
Harrell also participated in a study involving a brain-to-voice BCI, published in June 2025. While both BCIs use brain signals, this system goes beyond translating attempted speech into text, seeking to restore the expressiveness of natural speech. The study’s results suggest the future of communication for ALS patients could look remarkably different.
To make this possible, the researchers developed what the report calls a “real-time neural decoding pipeline” capable of synthesizing voice from incoming neural signals in under 10 milliseconds as a person attempts to speak. In other words, instead of waiting for an entire word or sentence to be decoded before generating speech, this model enables near-instantaneous speech synthesis more closely resembling the natural flow of speech.
In this study, researchers developed an AI-generated voice resembling Harrell’s using recordings of him speaking before ALS had severely impaired his speech. The BCI’s neural decoder—AI software that translates brain activity into information used to synthesize speech—allowed the system to produce speech from his attempted vocalizations.
The BCI also demonstrated success in reproducing his intended intonation. When he intended to ask a question, for example, the BCI reproduced the rising pitch associated with asking that question.
And this BCI demonstrated the ability to reproduce intended emphasis on words, too. Think about the difference in meaning between “I didn’t say she stole the money” and “I didn’t say she stole the money.” Emphasis may appear to be a small thing on the surface but can go a long way toward making day-to-day communication feel normal again.
And this study’s BCI takes speech restoration yet another step further, reproducing musical pitch. Harrell attempted to sing short melodies consisting of six or seven notes and successfully alternated between pitches classified as low, medium, and high.
The latter BCI study illustrates what is already being made possible by AI-powered research: real-time speech synthesis, intonation, emphasis, and even melody. In controlled trials, the system reproduced intended question intonation with 90.5 percent accuracy and intended word emphasis with 95.7 percent accuracy. Researchers also tested how well human listeners could distinguish musical pitches and understand the speech synthesized by the BCI.
When asked to identify which note was higher in 189 pairs of synthesized notes, listeners chose correctly 73.02 percent of the time, on average. When asked to identify the correct transcript of synthesized speech from six written sentence options, listeners chose correctly 94.34 percent of the time. Still, these results carry limitations.
Listeners were also asked to transcribe both the synthesized and unaided speech. Transcription, however, is a more demanding measure of intelligibility than choosing from six sentences. The median word error rate for the synthesized speech was 43.75 percent. And while this is a substantially lower error rate than when listeners were asked to transcribe Harrell’s unaided speech (96.43 percent), it remains a ways off from restoring natural conversation. These limitations underscore that while AI is already transforming communication for people living with ALS, accelerating research and innovation could bring us closer to restoring the flow of natural conversation and making these life-changing technologies widely available.
“I do not want to be special any longer,” Harrell wrote in a first-person account published by Springer Nature. “I think this is good enough technology that people who have ALS should have access to it right now. Sadly, that is not the case yet.” Sergey Stavisky, a neuroscientist at UC Davis who helped lead the BCI research, is optimistic about the future. He told Time, “I think motor-speech BCIs will be widely available.”
For patients living with ALS, advancements in AI and related emerging technologies are not merely about economic growth or productivity gains. They are about the promise of being able to communicate ordinarily again. And recent advancements in BCI technology illustrate both what AI is already making possible and what it can make possible if we keep going.
Harrell hopes to no longer be an exceptional case study, but one of many regaining the ability to communicate naturally, including those who may have believed they never would. Throughout history, extraordinary technological achievements have transformed what once seemed impossible into the commonplace. With enough research and innovation, Harrell’s dream may soon be in reach.
Ed Tarnowski is the Founder and Editor of The Compute Optimist. He is also a Resident Fellow, Technology and Innovation at the R Street Institute, a Senior Contributor at Young Voices, and the Washington, DC Regional Director for America’s Future. All opinions expressed are his own.