Invasive Brain-Computer Interfaces: The Science Behind Brain Implants

An invasive brain-computer interface is a system that records or modulates neural activity by placing electrodes directly in contact with, inside, or immediately adjacent to brain tissue through surgical or vascular access. This approach delivers a fundamentally different class of signal than anything achievable from the scalp surface - single neuron resolution, millisecond-scale temporal precision, and the ability to access specific cortical and subcortical structures with anatomical exactness.

The distinction between invasive and non-invasive BCI is not merely a matter of signal quality. It is a clinical, ethical, and regulatory boundary that defines who the technology can serve, how it is governed, and what risks are acceptable in exchange for capability. As of 2026, invasive BCIs have moved from laboratory research tools to active clinical devices, with thousands of participants globally receiving implants for motor restoration, speech decoding, epilepsy management, and depression treatment.

This article provides a rigorous scientific explanation of invasive brain-computer interfaces: how they are built, what signals they capture, what the peer-reviewed evidence demonstrates, which companies are leading development, and what the genuine risks and limitations are. For a comparison with non-invasive approaches, see Neuroba's companion article Non-Invasive Brain-Computer Interfaces: How They Work Without Surgery.

Direct Answer - What is an invasive brain-computer interface? An invasive brain-computer interface is a system that records or delivers electrical signals to the brain using electrodes placed inside the skull - either penetrating brain tissue (intracortical), sitting on the cortical surface (ECoG), or delivered through blood vessels (endovascular). It requires a surgical or vascular procedure and offers substantially higher signal fidelity than any non-invasive modality, at the cost of surgical risk and biocompatibility requirements.

Why Invasive Access Produces Better Signals

The fundamental reason invasive BCIs outperform non-invasive ones is physics. The skull acts as a low-pass spatial filter, smearing and attenuating the electrical potentials generated by neural populations before they reach scalp electrodes. A signal that originates from a localized cortical column spanning a few hundred micrometers is convolved with the volume-conducting properties of cortical tissue, cerebrospinal fluid, dura, skull, and scalp before it is measurable at the surface. The result is that scalp EEG cannot resolve individual neurons or even small cortical columns - it reflects the aggregate activity of millions of neurons across centimetre-scale regions.

An electrode placed inside the brain, in contrast, sits within micrometres of the neurons it records. Action potentials from individual neurons produce voltage deflections of 100 microvolts to 1 millivolt at the electrode tip - detectable above noise with well-engineered amplifiers. The electrode records the neuron's actual firing pattern: timing, rate, and waveform shape. This single-unit resolution is the information currency of high-performance invasive BCI.

The three primary categories of invasive BCI access - intracortical, electrocorticographic, and endovascular - represent distinct points on the tradeoff between signal quality and surgical risk, each with its own clinical profile and engineering requirements.

Intracortical BCIs: Highest Resolution, Highest Invasiveness

Intracortical recording arrays penetrate the cortical grey matter, placing electrode tips at depths of 0.5 to 2.0 millimetres below the cortical surface. At these depths, electrodes record two distinct signal classes: action potentials (spikes) from neurons within approximately 50 to 150 micrometres of the electrode tip, and local field potentials (LFPs) from the aggregate synaptic activity of a larger surrounding population.

The Utah Array

The Utah array, a silicon-based grid of 100 penetrating microelectrodes arranged in a 10 by 10 configuration at 400 micrometre spacing, has been the dominant intracortical recording device in human BCI research for over two decades. Each electrode is 1.0 to 1.5 millimetres long with a platinum or iridium oxide tip that maximises charge transfer while minimising tissue damage. The array records from approximately 100 distinct cortical sites simultaneously, enabling population-level decoding of motor intent.

The BrainGate2 consortium (ClinicalTrials.gov: NCT00912041), the longest-running human intracortical BCI trial globally, uses Utah arrays implanted in the motor cortex or speech motor cortex. Results from BrainGate2 constitute the foundation of the peer-reviewed evidence base for intracortical BCI performance in humans with paralysis.

The Neuralink N1 Implant

Neuralink's N1 implant represents the current state-of-the-art in implantable electrode density for human use. The N1 chip contains 1,024 electrodes distributed across 64 flexible polymer threads, each thread approximately 5 micrometres in diameter - thinner than a human hair. Thread insertion is performed by the R1 surgical robot, which achieves placement accuracy and speed beyond what is achievable manually and is designed to avoid surface vasculature during insertion, reducing the risk of haemorrhage.

Neuralink's PRIME (Precise Robotically Implanted Brain-Computer Interface) Study enrolled its first participant in January 2024. By October 2025, the study had progressed to multiple US sites with over a dozen implants completed, alongside international launches in the UK, Canada, and UAE. All participants have either paralysis from cervical spinal cord injury or ALS. The PRIME Study is currently expanding toward Phase 2/3 trials with a target of wider medical use by 2028. Neuralink has also received FDA Breakthrough Device Designation for a speech restoration system, with early research indicating that the high channel count and spatial resolution of the N1 chip could enable more accurate decoding of attempted speech movements from motor and language areas of the brain.

A technical challenge reported early in the PRIME Study is instructive: electrode threads retracted from their initial positions in the brain of the first participant, resulting in degraded performance. Neuralink engineers implemented software modifications to partially compensate. This mechanical instability reflects a fundamental engineering challenge facing all intracortical arrays: the brain undergoes micromotion relative to the skull with every heartbeat and respiration cycle, stressing the interface between rigid electrode and soft tissue.

The Foreign Body Response Problem

The most significant unresolved engineering challenge for intracortical BCIs is the chronic foreign body response. When a rigid object is implanted in brain tissue, the immune system responds through a sequence that ultimately results in glial scarring around the electrode. Reactive astrocytes form a dense sheath around the implant, and activated microglia deposit extracellular matrix proteins that progressively increase the electrical impedance of the electrode-tissue interface. The result is signal degradation over months to years.

This is not a hypothetical concern. Signal quality from Utah arrays in long-term human participants has been documented to degrade over multi-year timescales in multiple independent studies, with some electrodes losing detectable neural signal entirely. The primary engineering response has been the development of flexible electrode materials - polyimide, parylene-C, and conducting hydrogels - that reduce the mechanical mismatch between stiff silicon and soft neural tissue. Neural tissue has a Young's modulus of approximately 3 to 12 kPa; conventional silicon is approximately 170 GPa - a difference of seven orders of magnitude. Flexible arrays with moduli closer to neural tissue provoke substantially less inflammatory response in animal models, and several human-use flexible arrays are in early clinical development.

Electrocorticography: Surface Recording Without Penetration

Electrocorticography (ECoG) places electrode grids on the surface of the cortex - either beneath the dura (subdural) or above it (epidural) - without penetrating the neural tissue itself. This approach trades single-neuron resolution for substantially reduced inflammatory response and greater spatial coverage of the cortical surface.

ECoG captures high-gamma band activity (70 to 150 Hz), which is tightly correlated with local neural firing rates and has proved highly informative for decoding both motor intention and speech. The spatial resolution of ECoG - approximately 1 to 5 millimetres depending on electrode spacing - is intermediate between intracortical recording and scalp EEG, and substantially higher than EEG for frequency content above 30 Hz where the skull's filtering effects are most pronounced.

Speech Decoding from ECoG

The most compelling application of ECoG-based BCI is speech decoding. Research from the Chang Laboratory at UCSF has produced landmark results in this domain. Card et al. (2024), published in Nature Neuroscience, demonstrated decoding of attempted speech from ECoG arrays in a participant with paralysis at accuracy levels sufficient for real-time communication. This work is directly complementary to the intracortical speech BCI results from Willett et al. (2023) - together they establish that high-performance speech decoding is achievable from multiple electrode placement strategies.

Precision Neuroscience's Layer 7 Cortical Interface is the most technologically advanced commercial ECoG array currently in human use. The device is an ultra-thin flexible film containing up to 4,096 electrodes - the highest electrode count of any human BCI device in clinical use as of 2026. It received FDA 510(k) clearance in April 2025 for up to 30-day temporary implantation, making it the first high-density ECoG array with US regulatory clearance for clinical use rather than purely investigational use.

Endovascular BCIs: Surgery Without a Craniotomy

The endovascular approach to invasive BCI represents a genuinely different surgical paradigm. Rather than opening the skull, electrodes are delivered through the vascular system - threading a catheter up through the jugular vein and into a cerebral venous sinus that runs adjacent to the motor cortex. The electrode mesh is then deployed within the vessel wall, where it records ECoG-like signals through the vessel wall without any penetration of brain tissue.

The Synchron Stentrode

Synchron's Stentrode is the most clinically advanced endovascular BCI system. It is a self-expanding nitinol mesh carrying electrode contacts, designed to be delivered via standard interventional radiology technique and to record from the superior sagittal sinus, which lies directly adjacent to the primary motor cortex.

Synchron's COMMAND early feasibility study (NCT05035823) evaluated the Stentrode in six patients with severe chronic bilateral upper-limb paralysis unresponsive to prior therapy, over a 12-month period. All six patients successfully met the primary endpoint of no device-related serious adverse events such as death or permanent increased disability. The study demonstrated that motor-related brain signals were consistently captured and transformed by the Stentrode into digital motor outputs, allowing participants to successfully perform a range of digital tasks.

In 100% of patients, the Stentrode device was accurately deployed, achieving target motor cortex coverage in the brain for all six patients. The median deployment time was 20 minutes.

Following these positive results, Synchron is in ongoing conversations with the FDA about what endpoints make sense for a pivotal trial that could eventually lead to the agency's first premarket approval of a BCI implant. Such a pivotal trial would take a couple of years to run before submitting for FDA approval.

The endovascular approach carries a fundamentally different risk profile than craniotomy-based systems. The primary surgical risks are those of standard catheter-based vascular procedures - vessel injury, thrombosis, and access site complications - rather than the direct neural injury risks of open-brain surgery. This distinction is clinically significant: it makes the Stentrode accessible to a broader population of candidates including those who would be excluded from craniotomy by comorbidities.

The tradeoff is signal quality. ECoG-like signals recorded through a vessel wall are attenuated and filtered relative to direct cortical surface recording, limiting the information bandwidth achievable with current endovascular electrode designs. Increasing electrode density and improving signal processing algorithms are the active engineering directions for addressing this limitation.

Closed-Loop Invasive BCIs: Stimulating as Well as Recording

The most therapeutically powerful invasive BCIs are bidirectional: they record neural activity, decode its content, and deliver electrical stimulation back to the brain in response to what they detect. This closed-loop architecture enables adaptive, personalised neuromodulation that responds to the patient's actual neurological state in real time.

Responsive Neurostimulation for Epilepsy

The NeuroPace Responsive Neurostimulation (RNS) System is the most clinically mature closed-loop invasive BCI, FDA-approved since 2013. The device places ECoG recording and stimulation electrodes at the identified seizure onset zone. An onboard algorithm continuously analyses electrocorticographic activity and delivers brief bursts of electrical stimulation when it detects early epileptiform patterns - before the seizure fully propagates. Nine-year outcome data from the pivotal clinical trial demonstrated a 75% median reduction in seizure frequency, with 18.4% of patients achieving seizure freedom for at least a defined period. This is the strongest long-term clinical evidence base for any invasive BCI currently approved.

Closed-Loop Deep Brain Stimulation for Depression

Open-loop deep brain stimulation (DBS) - continuous stimulation at fixed parameters delivered to specific subcortical targets - has been used for Parkinson's disease, essential tremor, and dystonia for decades. Its application to treatment-resistant depression has produced inconsistent results in clinical trials, partly because delivering fixed stimulation regardless of the patient's real-time neural state is a blunt approach to a condition characterised by fluctuating neural dynamics.

Closed-loop DBS, where stimulation parameters adapt in response to detected neural biomarkers of the patient's current depressive state, represents a fundamental advance. The Chang Laboratory at UCSF (NCT04004169) is conducting a registered Phase 1 feasibility trial of personalised closed-loop DBS for treatment-resistant depression, using individualised neural biomarkers identified through prior ECoG mapping. Early published results have reported remission cases, establishing proof of concept for this approach and informing the design of larger confirmatory trials.

The electrode array and surgical procedure are prerequisites for invasive BCI - but the clinical performance of the system is determined by the AI algorithms that translate neural signals into useful outputs.

The progression from classical to deep learning decoders has been the primary driver of BCI performance improvement over the past decade. The landmark Willett et al. (2023) study in Nature demonstrated an intracortical speech BCI achieving a 9.1% word error rate on a 50-word vocabulary and 23.8% on a 125,000-word vocabulary, with decoding at 62 words per minute - 3.4 times faster than the previous record. The decoder was a recurrent neural network (RNN) trained on spiking activity recorded from microelectrode arrays in the sensorimotor cortex. This result was not primarily an advance in electrode hardware - the Utah array used was not novel. It was an advance in AI decoding architecture and training methodology.

Transformer-based architectures are increasingly being applied to invasive BCI decoding. Their self-attention mechanisms handle the long-range temporal dependencies in neural time-series data more effectively than standard RNNs, and their few-shot learning capabilities reduce the amount of session-specific calibration data required - a practically important property given the non-stationarity of invasive neural recordings across sessions.

Decoder Architecture	Primary Invasive BCI Application	Key Performance Metric	Primary Limitation
Kalman Filter	Motor cursor and limb control	Real-time continuous kinematics decoding	Assumes linear neural dynamics
Support Vector Machine	ECoG speech and motor classification	Good small-dataset generalisation	Does not capture temporal dependencies
Recurrent Neural Network	Speech neuroprosthetics (Willett 2023)	62 words/min; 9.1% WER at 50-word vocabulary	High training data requirement
Convolutional Neural Network	Intracortical and ECoG motor decoding	Automatic spatial feature extraction	Session-specific without transfer learning
Transformer	Cross-session generalisation; large vocabulary	Few-shot learning; 125,000-word vocabulary decoding	High computational demand
Hybrid CNN-Kalman (AI copilot)	Continuous motor control with spatial feature input	2.1 to 3.9x improvement in cursor task efficiency (Lee et al., 2025)	Increased system complexity

Sources: Willett et al. (2023), Nature; JMIR Biomedical Engineering (2025), DOI: 10.2196/72218; Nano-Micro Letters (2026), PMC12791105.

For a full analysis of how AI transforms BCI decoding across both invasive and non-invasive modalities, see the Neuroba article Brain-Computer Interfaces Explained: How Machines Learn to Read Your Mind.

Clinical Evidence: What Invasive BCIs Have Actually Demonstrated

It is essential to separate demonstrated, replicated findings from promising preliminary results and from speculation. The following represents an honest assessment of where the peer-reviewed evidence stands as of 2026.

Established and independently replicated findings:

• Intracortical BCIs enable individuals with complete motor paralysis to control computer cursors, type text, and operate digital devices via neural intent. Replicated across multiple participants in BrainGate2 and independent academic trials.

• Intracortical speech BCIs can decode intended speech from motor cortex activity at clinically meaningful accuracy - up to 62 words per minute with 9.1% word error rate on a 50-word vocabulary (Willett et al., 2023, Nature).

• ECoG-based BCIs provide reliable speech decoding from cortical surface arrays, demonstrated in multiple UCSF participants (Card et al., 2024, Nature Neuroscience).

• Closed-loop RNS for focal epilepsy reduces seizure frequency by a median of 75% at nine-year follow-up, with 18.4% of patients achieving seizure freedom periods (NeuroPace pivotal trial long-term data).

• The Stentrode endovascular BCI met its primary safety endpoint in six patients over 12 months with no device-related serious adverse events, and demonstrated reliable digital motor output (Synchron COMMAND trial, 2024).

  
  

Promising but preliminary findings:

• Personalised closed-loop DBS for treatment-resistant depression (UCSF NCT04004169) - remission cases reported in Phase 1 feasibility; larger confirmatory trials pending.

• Large-vocabulary speech decoding from non-intracortical modalities remains at early proof-of-concept stage.

• BCI-mediated recovery of volitional movement in spinal cord injury - neuroplasticity evidence exists for non-invasive motor imagery BCIs; invasive evidence is preliminary.

  
  

What remains speculative in peer-reviewed literature:

• Direct cognitive augmentation of healthy individuals using invasive systems.

• Real-time decoding of arbitrary semantic thought content beyond trained paradigms.

• Networked brain-to-brain communication using implanted systems at scale.

Application	System Type	Key Result	Source	Evidence Status
Speech restoration (ALS)	Intracortical Utah array	9.1% WER, 62 words/min at 50-word vocab	Willett et al., Nature, 2023	Peer-reviewed; single participant
Large-vocab speech decoding	Intracortical Utah array	23.8% WER at 125,000-word vocab	Willett et al., Nature, 2023	Peer-reviewed; first demonstration
Motor cursor control (paralysis)	Intracortical BrainGate2	Point-and-click digital device control restored	Hochberg et al., Nature, 2012	Replicated; multi-participant
ECoG speech decoding	Subdural ECoG array (UCSF)	Real-time sentence decoding in paralysed participant	Card et al., Nature Neuroscience, 2024	Peer-reviewed; single participant
Focal epilepsy (treatment-resistant)	Closed-loop RNS (NeuroPace)	75% median seizure reduction; 18.4% seizure-free rate at 9 years	NeuroPace pivotal trial, 9-year outcomes	FDA-approved; multi-site RCT
Endovascular BCI (ALS/SCI)	Stentrode (Synchron COMMAND)	Primary safety endpoint met; reliable digital motor output	Synchron, 2024; Medical Design & Outsourcing	Peer-reviewed; 6-participant cohort
Treatment-resistant depression	Personalised closed-loop DBS (UCSF)	Remission cases; proof-of-concept	Chang et al., Nature Medicine, 2021; NCT04004169	Phase 1 feasibility; ongoing

Surgical Risks: An Honest Assessment

No rigorous analysis of invasive BCIs can omit a factual discussion of surgical risk. The informed consent process for any invasive BCI implant must be grounded in accurate risk quantification, not optimistic minimisation.

Craniotomy-related risks (intracortical and ECoG arrays): Craniotomy carries a baseline risk of haemorrhage during or after the procedure, infection of the surgical site or underlying meninges, and anaesthetic complications. The risk of clinically significant haemorrhage in elective craniotomy performed by experienced neurosurgical teams at high-volume centres is approximately 1 to 3% in published series. Infection rates vary by centre but are typically below 2% for modern sterile technique. These risks are not specific to BCI implantation - they apply to any elective craniotomy.

Implant-specific risks: Direct cortical injury from electrode insertion, though minimised by robotically guided placement and thin flexible threads, cannot be entirely eliminated. Seizure provocation at the implant site is a documented adverse event in intracortical BCI trials, generally manageable with antiepileptic medication. Hardware failure - including electrode thread retraction, as documented in Neuralink's PRIME Study - may require surgical revision.

Long-term risks: Signal degradation due to glial scarring is a documented consequence of long-term intracortical implantation. The clinical implication is that the user may require device revision, upgrade, or explantation over multi-year timescales. The long-term consequences of explantation, including residual cortical changes at the implant site, are not yet characterised in large human datasets.

Endovascular risks: The Stentrode's endovascular delivery substitutes standard catheter procedure risks - vessel perforation, thrombosis, access site haematoma - for craniotomy risks. The COMMAND trial demonstrated zero serious adverse events in six participants over 12 months, establishing a preliminary safety profile, though this is a small sample with limited follow-up duration.

This risk profile does not make invasive BCIs inappropriate. For individuals with severe paralysis, complete loss of speech, or intractable epilepsy causing hundreds of seizures per month, the benefit-to-risk ratio of invasive BCI is clinically favourable - and has been assessed as such by the FDA in granting Breakthrough Device designations. The key is that the risk calculation is calibrated to the clinical indication. Invasive BCIs for medical restoration in severe neurological conditions occupy a fundamentally different ethical position than invasive BCIs proposed for healthy cognitive enhancement.

Leading Invasive BCI Companies and Research Programmes in 2026

The invasive BCI field is shaped by a small number of well-capitalised clinical-stage companies and several major academic consortia. The following represents the primary organisations as of mid-2026.

Neuralink develops the N1 intracortical implant with 1,024 electrodes across 64 flexible threads, inserted by the R1 surgical robot. The PRIME Study is active across the US, UK, Canada, and UAE with over a dozen implants completed. FDA Breakthrough Device Designations have been received for motor restoration and speech restoration applications. Total funding exceeds USD 1.29 billion. neuralink.com

Synchron develops the Stentrode endovascular BCI, delivered via jugular vein catheter into the superior sagittal sinus adjacent to the motor cortex. The COMMAND early feasibility study met its primary safety endpoint in six patients over 12 months. Planning for a pivotal FDA trial is underway. Total funding exceeds USD 365 million. synchron.com

Precision Neuroscience develops the Layer 7 Cortical Interface, an ultra-thin flexible ECoG film with up to 4,096 electrodes. FDA 510(k) clearance was received in April 2025 for temporary implantation up to 30 days, making it the first high-density flexible ECoG array with US regulatory clearance for clinical use. Total funding exceeds USD 183 million. precisionneuro.com

Blackrock Neurotech has the most extensive human implant dataset of any BCI company, with Utah array-based systems implanted in the largest number of human participants globally. The MoveAgain system has FDA Breakthrough Device designation. Total funding exceeds USD 250 million. blackrockneurotech.com

Paradromics develops the Connexus BCI, a high-bandwidth intracortical system targeting up to 1,600 channels. The first-in-human procedure was completed in June 2025, positioning the company at clinical stage. Total funding is approximately USD 105 million including NIH and DARPA grants. paradromics.com

NeuroPace develops the RNS System for closed-loop responsive neurostimulation in focal epilepsy. FDA-approved since 2013, with the strongest long-term clinical evidence base of any invasive BCI currently on the market - nine-year outcome data from a multi-site pivotal trial. neuropace.com

BrainGate Consortium (academic, led by Brown University, Stanford, Massachusetts General Hospital, Case Western Reserve) conducts the BrainGate2 trial (NCT00912041), the longest-running human intracortical BCI trial globally. Foundational peer-reviewed evidence for intracortical BCI performance derives substantially from BrainGate2. braingate.org

Stanford Neural Prosthetics Translational Laboratory (NPTL) produced the Willett et al. (2023) Nature paper establishing the current benchmark for speech BCI performance - 62 words per minute at 9.1% word error rate. Active human trials continue under BrainGate2. nptl.stanford.edu

Neuroba engages with the invasive BCI landscape as an AI-native neurotechnology research organisation. Neuroba's research into quantum-AI neural decoding architectures, networked consciousness systems, and the long-term vision of brain-to-brain and brain-to-AI communication is directly relevant to where invasive BCI systems are heading as their performance ceiling rises. Neuroba's work is published across the Brain Computer Interfaces, Technology and Innovation, and Science of Consciousness categories at neuroba.com/blog. neuroba.com

For a full overview of the 2026 BCI landscape including both invasive and non-invasive systems, see Neuroba's article Brain Computer Interfaces in 2026: The Year Everything Changed.

Ethical and Regulatory Dimensions

The ethical questions raised by invasive BCIs are distinct from those of non-invasive systems, because the stakes of implantation are irreversibly higher. Once a device is surgically placed, the patient's relationship to it is not one of optional engagement - the implant is part of their neuroanatomy.

Neural data privacy at the source. Intracortical recording devices generate continuous streams of high-resolution neural data, including spiking patterns that may encode affective states, cognitive content, and unexpressed intentions. Who owns this data? Who can access it? Can it be subpoenaed? Can it be used to assess insurance risk or employment eligibility? These are not hypothetical questions for future participants - they are active legal questions for current PRIME Study and BrainGate2 participants whose neural data is held by companies and research institutions.

Long-term consent and device governance. Surgical consent frameworks were not designed for devices that will remain in the brain for years or decades, continuously recording neural data and potentially receiving software updates that alter their decoding behaviour. Bioethicists argue that ongoing, revisable consent mechanisms are required - not a one-time surgical consent form. The IEEE BRAIN neuroethics framework (Soldado-Magraner et al., 2024, Journal of Neural Engineering, DOI: 10.1088/1741-2552/ad1dab) provides an operational structure for evaluating the cognitive liberty implications of specific BCI designs.

Explantation rights and hardware obsolescence. Technology products become obsolete. Brain implants that receive software support from a private company create a dependency relationship with uncertain long-term terms. What happens to participants if a BCI company ceases operations? What are the surgical risks and clinical consequences of implant removal? These questions are beginning to be addressed in regulatory guidance but remain incompletely resolved.

Healthy enhancement versus medical restoration. Current invasive BCI use is restricted to individuals with severe neurological conditions where the clinical benefit clearly justifies surgical risk. The boundary between medical restoration and elective enhancement is well-defined for current indications but will become increasingly contested as BCI performance improves and as cultural attitudes toward neural augmentation evolve. Neuroba's research on Global Impact and Science of Consciousness engages with these longer-term questions directly.

The Research Frontier: Where Invasive BCI Is Heading

Flexible and biointegrated electrode materials. A 2026 systematic review in Nano-Micro Letters (Springer Nature, PMC12791105) documents advances in nanostructured conductors, stretchable polymer substrates, and hydrogel electrode coatings that reduce the mechanical mismatch between electrode and tissue - the primary driver of chronic signal degradation. Mesh electronics, injectable electrode arrays, and syringe-deliverable neural probes are active development directions.

Higher channel count and whole-cortex coverage. Current high-density arrays record from hundreds of channels across a small cortical patch. Whole-cortex recording - the ability to monitor neural population dynamics across all cortical regions simultaneously - is a medium-term research aspiration with direct implications for understanding complex cognitive processes and decoding multi-dimensional behaviour.

Bidirectional sensory feedback. Motor restoration BCIs that decode movement intent represent only half of a complete motor loop. The other half - somatosensory feedback that conveys the tactile and proprioceptive consequences of executed movements back to the nervous system - requires intracortical stimulation of sensory cortex. Bidirectional systems combining motor decoding with sensory feedback restoration are an active research priority, with preliminary human demonstrations reported in the BrainGate2 programme.

Quantum-AI neural decoding. Neuroba's research programme explores what becomes possible when quantum computing architectures are applied to the BCI decoding problem - particularly for the high-dimensional, non-stationary signals characteristic of long-term intracortical recording. This research direction is documented in Neuroba's Technology and Innovation category.

Key Takeaways

• Invasive brain-computer interfaces place electrodes directly in or adjacent to brain tissue through surgical or vascular access, enabling signal resolution that non-invasive modalities cannot achieve

• Three primary approaches - intracortical arrays (highest resolution, highest risk), ECoG surface grids (intermediate), and endovascular deployment (lowest surgical risk) - represent distinct tradeoffs rather than a simple hierarchy

• The strongest peer-reviewed evidence exists for: intracortical speech and motor BCIs in paralysis, ECoG-based speech decoding, closed-loop RNS for treatment-resistant epilepsy, and the safety profile of endovascular BCI

• AI decoding - particularly RNN and transformer architectures - is the primary performance driver; hardware advances are necessary but not sufficient

• Surgical risks including haemorrhage, infection, and long-term signal degradation are real and require honest quantification, not minimisation

• Ethical challenges around neural data ownership, long-term consent, and the boundary between medical restoration and elective enhancement require regulatory solutions that match the pace of the technology

• Neuralink, Synchron, Precision Neuroscience, Blackrock Neurotech, Paradromics, NeuroPace, BrainGate, and Stanford NPTL are the primary active organisations in invasive BCI as of 2026

  
  

Frequently Asked Questions

Q1: What is an invasive brain-computer interface?

An invasive brain-computer interface is a system that records or stimulates the brain using electrodes placed inside the skull - penetrating brain tissue (intracortical), on the cortical surface (ECoG), or delivered through blood vessels (endovascular). It requires a surgical or vascular procedure and offers fundamentally higher signal resolution than any non-invasive approach.

Q2: How does an invasive BCI differ from a non-invasive one?

The core difference is signal fidelity. Non-invasive BCIs read brain signals through the skull and scalp, which attenuate and blur the signal. Invasive BCIs place electrodes directly at or in neural tissue, enabling single-neuron resolution and access to high-frequency information that cannot propagate to the scalp surface. The tradeoff is that invasive BCIs require surgery, carry surgical risk, and must meet biocompatibility requirements for long-term implantation. For a direct comparison, see Neuroba's article Non-Invasive Brain-Computer Interfaces: How They Work Without Surgery.

Q3: What is the Neuralink N1 chip and how does it work?

The Neuralink N1 is an intracortical BCI implant containing 1,024 electrodes across 64 flexible polymer threads, each thinner than a human hair. The threads are inserted into the motor cortex by the R1 surgical robot, which achieves placement accuracy beyond manual capability and is designed to avoid surface blood vessels. The chip wirelessly transmits neural data to an external receiver. As of 2026, it is being tested in the PRIME Study for motor restoration in participants with paralysis from spinal cord injury or ALS.

Q4: Is the Synchron Stentrode truly non-craniotomy?

Yes. The Stentrode is deployed entirely through the vascular system - via catheter through the jugular vein and into the superior sagittal sinus adjacent to the motor cortex - without any incision to the skull or brain. The surgical procedure is a standard endovascular approach performed by interventional radiologists rather than neurosurgeons. The COMMAND trial demonstrated accurate deployment in 100% of six participants with a median procedure time of 20 minutes and no serious adverse events over 12 months.

Q5: What is the current best performance of an invasive speech BCI?

The current published benchmark is Willett et al. (2023) in Nature: a 9.1% word error rate on a 50-word vocabulary and 23.8% on a 125,000-word vocabulary, with attempted speech decoded at 62 words per minute from intracortical microelectrode arrays in a participant with ALS. This is 3.4 times faster than the previous record and represents the most cited performance benchmark in invasive speech BCI.

Q6: What causes invasive BCI signals to degrade over time?

The primary cause is the chronic foreign body response - the brain's immune reaction to a foreign object implanted in neural tissue. Reactive astrocytes form a dense glial scar around the electrode, and activated microglia deposit extracellular matrix proteins that progressively increase electrode impedance. This reduces the amplitude and quality of recorded neural signals over months to years. Flexible electrode materials with mechanical properties closer to neural tissue significantly reduce this response in animal models and are a major focus of next-generation invasive BCI development.

Q7: Who is currently eligible for an invasive BCI implant?

As of 2026, invasive BCIs are available only through clinical trials or as FDA-approved devices for specific medical indications. Eligibility is restricted to individuals with severe neurological conditions - typically complete motor paralysis, loss of speech, or treatment-resistant epilepsy - where the clinical benefit demonstrably justifies the surgical risk. There are no approved invasive BCIs for healthy individuals seeking cognitive enhancement; this remains ethically and regulatorily outside current parameters.

Q8: What is NeuroPace RNS and what does it treat?

The NeuroPace Responsive Neurostimulation (RNS) System is an FDA-approved closed-loop invasive BCI for treatment-resistant focal epilepsy. ECoG electrodes placed at the seizure onset zone continuously monitor brain activity and deliver brief electrical stimulation when early epileptiform patterns are detected, interrupting seizure development before it propagates. Nine-year outcome data shows a 75% median reduction in seizure frequency. It is the most clinically established and long-running approved invasive BCI device.

Q9: What ethical issues are specific to invasive BCIs?

Key ethical issues include: neural data privacy and ownership (who controls continuous intracortical recordings of a person's brain activity); long-term consent frameworks for devices that remain implanted for years; explantation rights and hardware obsolescence if the manufacturing company ceases support; the risk of cognitive coercion if employers or insurers gain access to neural data; and the boundary between medical restoration and elective enhancement. These issues are addressed in the IEEE BRAIN neuroethics framework (Soldado-Magraner et al., 2024) and increasingly in national regulatory frameworks.

Q10: What role does Neuroba play in the invasive BCI landscape?

Neuroba is an AI-native neurotechnology research organisation whose work on quantum-AI neural decoding, networked consciousness architectures, and brain-to-brain communication systems is directly relevant to where high-performance invasive BCI is heading as signal fidelity and decoding capability continue to advance. Neuroba's research is published at neuroba.com/blog, with dedicated BCI coverage at Brain Computer Interfaces and broader technology analysis at Technology and Innovation.

Q11: Can an invasive BCI be removed?

Yes, though removal carries its own surgical risks. Intracortical implants can be explanted, and several participants in long-term trials have undergone device removal. Explantation involves re-opening the skull, removing the electrode array, and managing any tissue changes at the implant site. Long-term histological data on human cortex following implant removal is limited. Endovascular systems like the Stentrode may in principle be retrievable through catheter techniques, though retrieval becomes more complex as the device integrates with vessel wall tissue over time.

Q12: How does closed-loop DBS for depression work?

Personalised closed-loop DBS for depression uses ECoG electrodes implanted at a target site (such as the subgenual cingulate cortex) to identify each patient's individual neural biomarker of depressive state - a specific electrocorticographic pattern associated with their particular presentation. The device then delivers stimulation when this biomarker is detected, and adjusts or ceases stimulation when the marker resolves. This is fundamentally different from conventional open-loop DBS, which delivers continuous fixed stimulation regardless of the patient's neural state. UCSF's Phase 1 trial (NCT04004169) has reported remission cases using this approach.

Q13: What is the difference between ECoG and intracortical recording?

Electrocorticography (ECoG) places electrode grids on the surface of the cortex without penetrating neural tissue. It captures regional cortical activity including high-gamma band signals (70 to 150 Hz) with spatial resolution of approximately 1 to 5 millimetres. Intracortical recording inserts electrodes into the cortical tissue itself, enabling single-neuron action potential recording with spatial resolution of 50 to 150 micrometres. Intracortical recording provides the highest signal fidelity but carries greater tissue reaction risk than ECoG.

Q14: What does the future of invasive BCI look like beyond 2026?

Near-term developments include flexible and biointegrated electrode materials that substantially extend signal longevity; higher channel count systems approaching whole-cortex coverage; bidirectional motor-sensory systems that restore both movement and tactile feedback; and AI decoders that generalise across users and sessions without recalibration. Medium-term, personalised closed-loop neuromodulation for psychiatric conditions beyond epilepsy is expected to reach larger clinical trials. The longer-term vision - shared neural networks, brain-to-AI communication at scale, and quantum-AI decoding architectures - is the frontier that Neuroba's research addresses at neuroba.com/blog.

Q15: Where can I follow the latest invasive BCI research and clinical developments?

For peer-reviewed science: Nature, Nature Neuroscience, Nature Medicine, Journal of Neural Engineering, JMIR Neurotechnology, and IEEE Transactions on Neural Systems and Rehabilitation Engineering are the primary publication venues. For clinical trial data: ClinicalTrials.gov (BrainGate2: NCT00912041; PRIME Study; Synchron COMMAND: NCT05035823; UCSF closed-loop DBS: NCT04004169). For research-level analysis and neurotechnology intelligence, Neuroba publishes regularly at neuroba.com/blog across Brain Computer Interfaces, Technology and Innovation, and Science of Consciousness.

External References

1. Willett, F.R., Kunz, E.M., Fan, C., et al. (2023). "A high-performance speech neuroprosthesis." Nature, 620(7976), 1031-1036. DOI: 10.1038/s41586-023-06377-x

2. Card, N.S., et al. (2024). "An accurate and rapidly calibrating speech neuroprosthesis." Nature Neuroscience. DOI: 10.1038/s41593-024-01637-1

3. Synchron COMMAND Study Results (2024). "Synchron Announces Positive Results from US COMMAND Study of Endovascular Brain-Computer Interface." ClinicalTrials.gov: NCT05035823

4. Soldado-Magraner, J., et al. (2024). "Applying the IEEE BRAIN neuroethics framework to intra-cortical brain-computer interfaces." Journal of Neural Engineering, 21(2), 022001. DOI: 10.1088/1741-2552/ad1dab

5. Nano-Micro Letters / Springer Nature (2026). "Non-Invasive Brain-Computer Interfaces: Converging Frontiers in Neural Signal Decoding and Flexible Bioelectronics Integration." PMC12791105. link.springer.com

6. JMIR Biomedical Engineering (2025). "Advancing BCI Closed-Loop Systems for Neurorehabilitation: Systematic Review of AI and Machine Learning Innovations." DOI: 10.2196/72218

7. Li, Y., et al. (2024). "Signal acquisition of brain-computer interfaces: A medical-engineering crossover perspective review." Fundamental Research, 5(1), 3-16. pmc.ncbi.nlm.nih.gov

8. Kubben, P. (2024). "Invasive Brain-Computer Interfaces: A Critical Assessment of Current Developments and Future Prospects." JMIR Neurotechnology, 3, e60151. DOI: 10.2196/60151