Brain-Computer Interfaces Explained: How Mind-Controlled Tech Actually Works
In early 2024, a paralysed man moved a computer cursor with his thoughts — and then stayed up late playing chess online with it. The chip in his head was built by Neuralink, and the moment lit up the internet as the beginning of the "mind-reading" era. But brain-computer interfaces (BCIs) are neither mind-reading nor magic. They're pattern-recognition machines plugged into the most complex pattern generator we know.
Here's what they are, who is building them, and what they honestly can and can't do.
What a BCI actually does
A brain-computer interface is a system that records electrical activity from the brain and translates it into commands a computer can understand. The most practical targets are the motor cortex — the brain region that issues movement instructions.
The pipeline is surprisingly straightforward:
- Record. Electrodes pick up the tiny voltage spikes (action potentials) that neurons fire when you intend to move.
- Decode. Machine-learning software learns the pattern: this cluster of firing means "move the cursor left", that burst means "click".
- Act. The decoded intent drives a cursor, a robotic arm, a speech synthesiser, or a wheelchair.
The crucial point: the device reads motor intent, not thoughts. It can tell you're trying to move your right hand because your motor cortex fires as if to move it — even when the spinal cord can't deliver the message to the body. It cannot tell what you're dreaming about, what you believe, or what you had for breakfast. "Mind control" is a fair description of the cursor; "mind reading" is not.
The three lanes of the race
No single design is winning outright. The field has split into lanes that trade bandwidth (how much signal you capture) against surgical risk. As of October 2026, no fully implanted BCI for paralysis has full US FDA approval — every device is still investigational.
| Company | Approach | Electrodes | Human implants | Regulatory status |
|---|---|---|---|---|
| Neuralink (N1) | Ultra-thin threads inserted into the cortex by a surgical robot | ~1,024 | 20+ in the PRIME trial (first implant Jan 2024) | Investigational; two more trials (CONVOY for robotic arms, VOICE for speech) |
| Synchron (Stentrode) | Mesh device delivered through a blood vessel — no open-brain surgery | 16 | ~10 patients | First-ever IDE for a chronically implanted BCI; pivotal trial planned toward a first-of-its-kind approval |
| Paradromics (Connexus) | Fully implanted intracortical chip, speech-focused | 400+ | First permanent implant 2026 | FDA IDE granted Nov 2025 |
| Precision Neuroscience (Layer 7) | Thin-film array laid on the brain's surface | ~4,096 across 4 arrays | Dozens tested (short-term) | 510(k) clearance Apr 2025 for up to 30-day temporary use |
The trade-off is the story. Neuralink's penetrating threads sit among the neurons themselves, so the signal is rich — rich enough, reportedly, for fast cursor control and gaming. Synchron's Stentrode trades that resolution away: it's guided through the jugular vein to the motor cortex in a procedure similar to a cardiac stent placement — about 20 minutes, no skull opened — and reads signals through the vessel wall. Less bandwidth, dramatically less surgery. Precision sits in between: thousands of electrodes on a film resting on the brain's surface, high channel count without penetrating tissue.
Which lane "wins" will probably depend on the patient: someone who can't tolerate brain surgery fits Synchron's endovascular route; an ALS patient who has lost speech and needs fluent communication fits the high-bandwidth implants.
What they can do today
The results are real but narrow — and that's worth stating plainly.
- Cursor and device control. Paralysed participants move cursors, browse the web, and play games at speeds approaching a normal mouse.
- Speech restoration. Devices decode attempted speech into text or synthesised voice — transformative for people with ALS who have lost the ability to speak.
- Robotic limbs. Neuralink's CONVOY trial is testing thought-controlled robotic arms; research systems have let participants feed themselves.
- Communication for locked-in patients. Synchron's earliest work focused on people who cannot move or speak at all, giving them a digital channel to the world.
A 2026 landscape study counted only a few dozen implantable BCI trials ever run, involving a small number of participants in total. This is still early-stage medicine, not a consumer product. There are no commercially available BCIs you can buy, and none is approved in the US or EU for general use.
The honest limits and risks
Surgery has risks. Opening the skull (or even threading a device through a vein) carries infection and bleeding risks like any neurosurgery. Invasive electrodes can also degrade: scar tissue forms around implants over months and years, and signal quality can fade — one of the field's hardest unsolved problems.
"Mind reading" is a myth — for now. Today's devices decode the intention to move. Researchers are decoding more complex things, like attempted speech, but a BCI that reads your inner monologue does not exist. Still, ethicists are already asking the right questions early: who owns your neural data? What happens if a company with your brain implant goes bankrupt? These aren't hypotheticals forever — they're the terms-of-service questions of the 2030s.
Hype outpaces evidence. Company announcements move faster than peer-reviewed results. Several impressive demos come from company blogs and press releases, not independent studies. The rule that applies to all medical technology applies here: a press release is not a clinical result.
Takeaway checklist
- What it is: a BCI records electrical activity from the brain — usually the motor cortex — and machine learning decodes the intent to move into computer commands.
- What it's not: mind reading. It decodes movement intent, not thoughts, memories, or beliefs.
- The approaches: Neuralink's invasive threads (highest bandwidth), Synchron's blood-vessel Stentrode (no open-brain surgery, lower resolution), Precision's surface film (middle ground), Paradromics' speech-focused implant.
- The state of play: dozens of participants across clinical trials; Synchron raised $200M in late 2025 and is planning a pivotal trial; Neuralink raised $650M in a Series C in June 2025 — but no fully implanted BCI has full FDA approval as of October 2026.
- The risks: neurosurgery carries real risks, implants can degrade over time, and neural data raises ownership and privacy questions we haven't solved.
- Who it's for today: people with paralysis, ALS, and locked-in syndrome — not consumers. The first real products will be medical devices, not gadgets.
Brain interfaces are one frontier of medical technology; another is already inside millions of people. Read our guides to CRISPR and gene editing, explained and the growing role of AI in healthcare for the rest of the story.
