Brain-Computer Interface 2026: What Neuralink's Human Trials Have Shown
In January 2024, Noland Arbaugh — paralysed from the shoulders down after a diving accident — had a chip implanted in his brain by Neuralink. Within weeks, he was playing chess online and browsing the internet using only his thoughts. No mouse. No keyboard. No physical movement at all.
This is not science fiction. Brain-computer interfaces (BCIs) are real devices in human clinical trials right now, and the results from those trials are changing what we understand about the boundary between human cognition and digital systems. Neuralink is the most famous company in this space, but it is far from the only one — and not necessarily the most clinically advanced.
This guide explains how BCIs work, what Neuralink's human trial actually showed, who the other serious players are, what BCIs can currently do for patients with paralysis, what remains unsolved, and the ethical questions that the technology forces into the open.
How Brain-Computer Interfaces Work
A brain-computer interface reads electrical signals fired by neurons in the brain, converts them into digital data using signal processing algorithms, and translates that data into commands for external devices. The result: direct control of computers, robotic limbs, or communication tools using thought alone.
Neurons communicate through electrical signals. When you think about moving your hand, motor neurons in your brain fire in specific patterns. A BCI detects those patterns by placing electrodes — thin wires or arrays — near or inside the brain tissue. The electrodes pick up the electrical activity and send it to a signal processor.
The signal processor uses machine learning algorithms to decode the neural firing patterns into digital commands. Early BCIs required weeks of calibration — the patient would think about movements while the algorithm learned what those patterns meant. Modern systems learn faster and adapt in real-time as patterns drift over days or weeks.
The output of that decoding — a cursor position, a letter selection, a robotic arm movement — is then sent to the external device. The whole loop from neural signal to device response currently takes around 50–200 milliseconds in the best systems, which is fast enough to feel responsive.
The Core Engineering Challenge
The brain has approximately 86 billion neurons. Current implanted BCI systems read from a few hundred to a few thousand electrodes simultaneously — a tiny fraction of the total. The field's fundamental challenge is reading more neurons with higher signal-to-noise ratio, from devices that last decades inside human tissue without degrading.
Neuralink's N1 chip uses 1,024 electrodes — more than most competing devices. But even 1,024 electrodes capture only a small slice of the neural activity relevant to complex movements or thoughts. The signal processing AI compensates by learning which patterns matter, but the resolution is still limited compared to what the brain actually generates.
Neuralink's Human Trials — What Actually Happened
Neuralink received FDA approval for its PRIME study (Precise Robotically Implanted Brain-Computer Interface) in May 2023. The first human implantation took place in January 2024. The patient, Noland Arbaugh, has ALS-related paralysis from a diving accident. He was 29 years old at the time of implantation.
Within 23 days of surgery, Arbaugh was using the device. He demonstrated cursor control, played chess online, and played the strategy game Civilization VI — all through thought. His cursor movement speed was reported to exceed the previous record for BCI cursor control set by BrainGate research participants, at over 8 bits per second of information transmission.
A second participant was implanted in 2024. Neuralink's stated goal for the PRIME study is to demonstrate that paralysed people can control external devices — computers, smartphones, robotic limbs — using the implant. The company aims to eventually enable movement restoration through a second device, CONVOY, that would stimulate spinal cord neurons below the injury site.
Other BCI Companies: Synchron, BrainGate, Blackrock
Synchron — No Open Brain Surgery Required
Synchron is developing the Stentrode — a BCI device implanted via a blood vessel rather than through open brain surgery. A surgeon threads the device through the jugular vein into a blood vessel near the motor cortex. No drilling into the skull. This makes the procedure far less risky and suitable for patients who cannot safely undergo open neurosurgery.
The Stentrode has been implanted in patients in Australia and the US. Published results show ALS patients able to control computers and send messages using the device. Signal resolution is lower than directly implanted electrode arrays, but the safety profile is significantly better — making it a credible clinical alternative for broader deployment.
BrainGate — The Academic Pioneer
BrainGate is a research consortium (Brown University, Massachusetts General Hospital, and others) that has been implanting BCIs in human patients since 2004. Their research participants — primarily people with ALS and spinal cord injuries — have demonstrated robotic arm control, cursor movement, and speech synthesis from neural signals. BrainGate's work established most of the foundational science that commercial companies now build on.
Blackrock Neurotech
Blackrock Neurotech has implanted its Utah Array BCI device in over 40 patients across multiple research programs — more human implantations than any other company as of 2024. Their devices have enabled patients to feel sensations through robotic hands and to control robotic limbs with realistic dexterity. Unlike Neuralink, Blackrock's primary focus is research and medical applications rather than consumer products.
Precision Neuroscience
Precision Neuroscience is developing a thin-film electrode array — a device so thin it can slide under the skull without penetrating brain tissue. This "epidural" approach sits between the skull and the brain surface, reading signals with less invasiveness than depth-implanted electrodes. Clinical trials are in early stages but the company has placed the device temporarily in neurosurgery patients to demonstrate signal capture capability.
Medical Uses: Paralysis, ALS, and Restoring Communication
The near-term medical applications of BCIs are already changing patients' lives:
- Cursor and computer control for paralysis: Patients with quadriplegia, ALS, or locked-in syndrome can use BCIs to control computers, send emails, and browse the internet — restoring meaningful independence and communication.
- Robotic limb control: BrainGate and Blackrock research has demonstrated that patients can control robotic arms with enough precision to pick up objects, shake hands, and perform everyday tasks.
- Speech synthesis: Research teams have decoded neural signals associated with attempted speech from non-verbal patients, enabling a form of direct thought-to-text communication. A 2023 Nature paper demonstrated a paralysed patient producing synthesized speech at 78 words per minute from neural signals.
- Seizure detection and treatment: Non-implanted and semi-implanted neural monitoring devices are already FDA-cleared for detecting epileptic seizures and triggering responsive neurostimulation to interrupt them.
- Depression treatment: Deep brain stimulation (DBS) — an established neurosurgical procedure — uses implanted electrodes to modulate circuits involved in treatment-resistant depression. Next-generation closed-loop DBS systems use AI to adapt stimulation based on real-time neural feedback.
Invasive vs Non-Invasive BCIs
Not all BCIs require surgery. The tradeoff is clear: less invasive means lower signal quality.
| Type | Examples | Signal Quality | Risk Level | Current Use |
|---|---|---|---|---|
| Fully invasive (depth implant) | Neuralink N1, Utah Array | Very high (single-neuron) | High (brain surgery) | Clinical trials, research |
| Semi-invasive (epidural/endovascular) | Synchron Stentrode, Precision | Medium-high | Medium (vascular surgery) | Clinical trials |
| Non-invasive (EEG headset) | Neurosity Crown, OpenBCI | Low (population signals) | Zero | Consumer / research |
Consumer EEG headsets exist and can detect broad mental states — focused vs relaxed, for example — but they cannot read the fine-grained neural patterns needed for precise device control. The signal-to-noise ratio from electrodes on the scalp, filtered through the skull, is far too low for the applications that make BCIs clinically meaningful.
Future Possibilities Beyond Medical Use
Beyond restoring lost function, longer-term BCI applications being researched include:
- Memory augmentation: DARPA-funded research has shown that neural stimulation can improve memory formation in people with traumatic brain injury. The same principle could theoretically enhance memory in healthy individuals.
- Direct brain-to-brain or brain-to-AI communication: Research demonstrations have achieved rudimentary brain-to-brain information transfer via BCI — one person imagines a signal, it is decoded and transmitted, another person receives a stimulation. This is extremely limited today but theoretically extensible.
- Sensory restoration and enhancement: BCIs combined with artificial retinas are restoring partial vision to blind patients. The same logic applies to hearing, touch, and proprioception.
- Emotion regulation: Closed-loop neurostimulation that detects anxiety or depression states and modulates the relevant brain circuits in real-time is in early research phases.
Neuralink's public statements describe a long-term goal of merging human cognition with AI — the ability to directly interface with large language models and AI systems at the speed of thought, rather than through typing or speech. This remains many decades of engineering away from the current state of the technology.
Ethical Concerns — The Hard Questions
BCIs raise ethical questions that existing frameworks for medical devices or AI do not fully address:
Mental privacy: If a device reads your neural signals, it potentially reads your thoughts — not just your intended movements. Who owns that data? What prevents a BCI company from analysing thought patterns for commercial or government purposes? No comprehensive legal framework yet protects neural data specifically.
Cognitive enhancement inequality: If BCIs enable enhanced memory, faster processing, or direct AI access for people who can afford them, this creates a new dimension of inequality — between cognitively augmented and non-augmented people. Unlike most technologies, this advantage would be literally built into the brain.
Corporate control of neural infrastructure: Neuralink is a private company. If a patient has a Neuralink implant and the company fails, changes its terms of service, or is acquired by another entity, what happens to the device? Medical device companies have obligations that software companies do not — but the regulatory landscape for neural data is still being written.
Identity and agency: A device that modulates brain activity raises questions about what constitutes autonomous thought and decision-making. If a closed-loop system alters your emotional state based on neural readings, is the resulting decision still fully yours? These are not hypothetical — they apply to current-generation deep brain stimulation devices.
Consent in vulnerable populations: Many BCI patients have conditions — ALS, locked-in syndrome, severe paralysis — that make them desperate for any quality-of-life improvement. The power imbalance between trial participants who have few options and companies seeking to demonstrate feasibility creates complex consent dynamics.
The rapid advance of AI capabilities described in our overview of Google I/O 2026 AI developments is converging with BCI hardware in ways that will accelerate these questions. For businesses thinking about the AI frontier more broadly, our guide to AI agents in 2026 covers how AI decision-making is already changing industries today.
