Neural Interfaces

 Brain Computer Interfaces

Brain–computer interfaces (BCIs) create a direct communication pathway between neural activity and an external device. The most credible near-term goal is not science-fiction “mind reading”, but restoration: giving people with paralysis or loss of speech a new route to communicate, use computers and, increasingly, control assistive devices.

What a BCI actually does

A BCI records patterns of neural activity, processes those signals with decoding algorithms and maps the decoded intent onto an output. Depending on the system, the output might be cursor movement, selection of letters, control of a smartphone, operation of an assistive robotic device, or synthesized speech. Some systems can also stimulate neural tissue, but most of today’s headline communication BCIs are primarily focused on recording and decoding.

Why 2025–2026 changed the picture

The field has moved from isolated laboratory demonstrations toward early clinical ecosystems. The important change is not that BCIs are suddenly routine treatments—they are not—but that multiple platforms are now being tested in people, at home as well as in research settings, and are targeting practical functions rather than only laboratory benchmarks.

• Neuralink expanded beyond the United States. GB-PRIME launched in Great Britain in 2025, with UCLH and Newcastle as study sites.
• At UCLH, the first UK participant was able to begin controlling a computer cursor by thought the day after implantation in October 2025.
• By 29 January 2026, UCLH reported seven GB-PRIME participants had undergone Neuralink implant surgery between October and December 2025.
• Paradromics began its FDA-authorised Connect-One early feasibility study and completed the first long-term Connexus implantation in June 2026.
• Synchron continues to pursue a less invasive endovascular BCI delivered through the jugular vein rather than by open cranial implantation.
• Speech BCIs took a major step forward in 2025, with systems demonstrating continuously streamed synthetic speech and more expressive brain-to-voice output.

Neuralink: where it stands now

Neuralink’s N1 system is a fully implantable, wireless intracortical BCI. In the Great Britain study, the implant uses more than 1,000 electrodes distributed across ultra-fine threads placed into targeted motor-cortex tissue by the R1 surgical robot. The system records neural activity associated with movement intention and uses software to convert it into cursor movements, clicks and keyboard actions.

Beyond Neuralink: the leading approaches

The biggest scientific leap: restoring speech

Speech neuroprostheses may ultimately have a more profound clinical impact than cursor control. Instead of asking a person to select letters one at a time, these systems decode neural activity generated during attempted speech and translate it into text or audible voice.

This matters because natural conversation depends on timing, prosody and identity—not just accurate text. The emerging goal is therefore a neural “voice box”: a system that can restore not only the words a person intends to say, but increasingly the pace and expressive qualities of speech.

Four broad BCI strategies

1. Intracortical interfaces — Electrodes penetrate the cortex and can capture very high-resolution neural activity. This can support fine motor decoding and speech research, but requires neurosurgery and raises questions about long-term tissue response, signal stability and device maintenance.
2. Cortical-surface interfaces (ECoG / micro-ECoG) — Electrodes rest on the brain’s surface rather than penetrating deeply. They can provide rich signals across a larger cortical area with less direct tissue penetration, although implantation still generally requires a neurosurgical procedure.
3. Endovascular interfaces — Electrode systems such as Synchron’s Stentrode are delivered through blood vessels and positioned near motor cortex. The attraction is reduced surgical invasiveness; the trade-off is that the electrodes are farther from individual neurons.
4. Non-invasive BCIs — EEG, MEG and emerging optical or ultrasound approaches avoid implantation. They are easier to deploy but generally have lower signal-to-noise ratio and spatial resolution, limiting the bandwidth available for fast, complex control.

What BCIs can do now — and what they cannot

Safety, durability and the difficult questions

The central challenge is no longer simply “can a neural signal control a computer?” The harder translational question is whether an implant can remain safe, useful, stable and supportable for years across a much wider range of people.

• Surgical and medical risk: Infection, bleeding, seizures, vascular injury, device migration, scarring and other procedure-related complications vary with the implantation route.
• Signal longevity: Neural signals can change over time because of tissue response, electrode movement, disease progression or changes in neural activity. A 2024 NEJM report described seven years of successful implanted BCI use in ALS before progressive signal decline ultimately made the system unreliable.
• Cybersecurity: Wireless implants and connected software create a new safety domain: device authentication, secure updates, adversarial access and protection against manipulation of neural data.
• Mental privacy: Brain data may reveal highly sensitive information about intention, attention, movement preparation and, in some contexts, aspects of speech. Governance needs to treat neural data as unusually sensitive.
• Agency and consent: Users need meaningful control over what is decoded, stored, transmitted or acted upon. The distinction between assisting a user and algorithmically interpreting a user becomes ethically important.
• Access and inequality: If advanced neurotechnology works but remains extremely expensive or concentrated in a few specialist centres, the benefits could be distributed very unevenly.

Neural privacy: a new category of personal data

Neuroethics is moving from philosophy into practical regulation. International work by organisations including the OECD and UNESCO has emphasised mental privacy, brain-data protection, informed consent, autonomy and safeguards against misuse. As BCIs become more capable, these principles will matter as much as electrode count or decoding speed.

What could happen next?

2026–2028 — More participants, longer follow-up and increasingly independent home use. Expect competition around surgical simplicity, signal quality, wireless reliability and software usability.

Speech restoration — Brain-to-voice systems are likely to become a major clinical target for ALS, brainstem stroke and severe paralysis, with emphasis on conversational speed, personalised voice and low training burden.

Assistive robotics — BCI control may expand from screens to robotic arms, powered wheelchairs and smart-home systems, but reliable safety-critical control is a higher bar than moving a cursor.

Bidirectional systems — Longer-term systems may combine recording with stimulation, opening routes to sensory feedback—for example, delivering tactile information from a prosthetic hand back into the nervous system.

AI-assisted decoding — Machine learning and language models can improve prediction, adaptation and error correction. The challenge is to make AI assist the user without silently substituting the model’s own intent.

Consumer enhancement — Healthy-person enhancement remains much more speculative. Clinical benefit for severe disability is the realistic frontier; memory uploading, superhuman cognition and mass-market neural–AI fusion remain unproven.

Selected current sources

1. University College London Hospitals NHS Foundation Trust. Seven GB-PRIME patients now participating in Neuralink trial. 29 January 2026. Open source
2. University College London Hospitals NHS Foundation Trust. First UK patient uses thought to control computer hours after Neuralink implant. 27 October 2025. Open source
3. Neuralink. GB-PRIME Study Launch. 31 July 2025. Open source
4. UCLH. GB-PRIME study overview and N1 system details. 31 July 2025. Open source
5. Synchron. Technology and active clinical studies. Accessed August 2026. Open source
6. Paradromics. First Connexus BCI implantation in the FDA-approved Connect-One study. 17 June 2026. Open source
7. U.S. Food and Drug Administration. 510(k) K242618, Layer 7-T cortical electrode, Precision Neuroscience. Decision 30 March 2025. Open source
8. Littlejohn KT et al. A streaming brain-to-voice neuroprosthesis to restore naturalistic communication. Nature Neuroscience. 2025. Open source
9. Wairagkar M et al. An instantaneous voice-synthesis neuroprosthesis. Nature. 2025;644:145–152. Open source
10. Card NS et al. An Accurate and Rapidly Calibrating Speech Neuroprosthesis. New England Journal of Medicine. 2024;391:609–618. Open source
11. Vansteensel MJ et al. Longevity of a Brain–Computer Interface for Amyotrophic Lateral Sclerosis. New England Journal of Medicine. 2024;391:619–626. Open source
12. NIH. Brain-computer interface restores natural speech after paralysis. 2025. Open source

• Clinical status note: The implantable BCI systems described here remain investigational unless a specific regulatory clearance is stated. Regulatory clearance of a component does not mean that a complete long-term BCI system is approved for routine clinical use.