brain-computer interface (BCI)

basicexperimental

A tool that measures brain activity, typically its electrical signals, and translates them into machine commands, allowing control of a computer or other device by thought, without muscle movement.

In plain terms

A brain-computer interface is like a translator between your brain and a machine. When you want to move your hand, countless neurons in your brain send tiny electrical signals. The interface monitors these signals and tries to figure out what you want to do.

Today it is primarily a medical device. It helps paralyzed people move a cursor with their thoughts, write text, control a robotic arm, or speak again in a synthetic voice after losing the ability to speak.

It differs from the mouse, keyboard, or voice control in that it requires no muscle: neither hand nor mouth. Signals are currently typically deciphered by artificial intelligence, which is why the field has accelerated along with AI development.

The vision is bolder: even healthy people could assign tasks directly with their thoughts to a machine, for example, an AI agent. This is currently a plan, not a finished product.

How it works

Sensing: electrodes measure the brain's electrical activity. These can be on the scalp (EEG cap), on the brain's surface, implanted into brain tissue, or delivered near the brain through a vein.
Signal cleaning: raw signals are very noisy, so the system filters out interfering parts, such as noise from muscle movement or surrounding electrical devices.
Decoding: a machine learning model learns which signal pattern corresponds to which intent in your case, for example 'cursor left' or the pattern of a sound you are trying to say.
Execution: the recognized intent becomes a command, meaning the cursor moves, a letter appears, a machine voice speaks, or a robotic arm moves.
Feedback and learning: you see the result, and both you and the model gradually adapt to each other. This is why the tool must be trained and occasionally retuned.

An everyday example

In the news, you see a man paralyzed from the neck down playing chess on his laptop, even though his hands do not move. A coin-sized device implanted in his brain reads where he wants to move the cursor. A more everyday version available in stores is an EEG headband that tries to measure your attention or sleep, but this is much less accurate.

Who is affected

patient, doctorFor those living with paralysis, ALS, or speech loss after a stroke, this can restore the possibility of communication and independent device use. → Health

lawyer, citizenBrain data is one of the most personal types of data, raising new questions about who can access it and how it can be used. → Law

developerDeciphering signals is a machine learning task, and brain-signal-controlled applications open a new development area. → Code

everyoneSome consumer headbands and headphones already collect brain signals today, so it is worth knowing what they actually measure.

A common misconception

MythThe brain-computer interface reads your thoughts.

In realityToday's devices cannot read your secret thoughts. They recognize a few trained intentions, such as the pattern of a movement or of a word the user is trying to say, and the device must be tuned to the user. However, research has already managed to detect fragments of inner speech, so the protection of brain data is a real issue.

Limits and risks

The most precise tools require brain surgery, which carries risks of infection and bleeding, and electrode performance may degrade over time.
Non-invasive devices that measure from the scalp, without surgery, pick up a weak, noisy signal, so they can do far less than their advertising suggests.
Brain data can be used to infer health status, attention and mood, so check where it goes and who handles it.
It remains an open question what will happen to the implanted device and the patient if the manufacturer ceases operations or stops support.

Similar, but not the same

EEG examination – The EEG only measures and records brain signals for diagnostic purposes, while the interface also turns them into real-time commands.
deep brain stimulation – A stimulator delivers current to the brain, for example for Parkinson's disease treatment, and does not read your intentions for device control.
voice control – Voice control requires you to speak the command, whereas the interface works directly from brain signals without speech or movement.

Related concepts

Broader

human-machine interface, neurotechnology

Narrower

brain implant, speech neuroprosthesis, non-invasive brain-computer interface

On the edge

machine learning, neural data, augmented reality, wearable device, AI-agent

A brief history

The electrical signals of the brain were first recorded in humans by the German psychiatrist Hans Berger in 1924. The term "brain-computer interface" was introduced by Jacques Vidal, a researcher at the University of California (UCLA), in a 1973 study. Experiments on humans with implanted devices have been conducted since the early 2000s (for example, in the BrainGate studies). Neuralink implanted its first device into a human in January 2024, bringing the topic to a wide audience.

Where things stand

At the end of 2026, implantable devices are operating in clinical trials with few patients, in competition among several companies (such as Neuralink and Synchron). Off-the-shelf, non-surgical headbands exist, but their capabilities are limited, and thought control for healthy people is still a vision of the future.

Status: 2026 November

Also known as

brain-computer interface · brain–computer interface · brain-computer connection · brain interface · BCI · brain-computer interface · brain-machine interface · BMI

Sources

Wikipedia: Brain–computer interface

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.