Three BCI Architectures, One Regulatory Finish Line: What the 2025 Trial Data Actually Shows

6 min read

Key Claim: Three companies have produced clinical BCI safety data in the past 18 months using fundamentally different architectures, regulatory strategies, and implant methods — the divergence is not incremental refinement but competing bets on which approach survives the approval process.

In late 2024, Synchron presented the 12-month safety results from its COMMAND study: six patients, zero serious adverse events, 100% device deployment success — all without a single craniotomy. That same autumn, Precision Neuroscience published its first peer-reviewed clinical data in Nature Biomedical Engineering. By early 2025, Neuralink had crossed 4,900 cumulative hours of device use in its PRIME cohort. Three companies, three architectures, and — critically — three different bets on how to get a brain-computer interface through FDA premarket approval. None of them has done it yet.

The Architecture Question the Coverage Keeps Missing

The phrase “brain implant” obscures a decision that determines almost everything downstream: where does the electrode sit relative to cortical tissue?

Neuralink’s N1 implant is intracortical. Its 1,024 electrodes, distributed across 64 flexible threads thinner than a human hair, penetrate the cortical surface and sit directly among neurons. This maximizes signal resolution — single-unit activity, the gold standard for neural decoding — but it requires a robotic surgical procedure and carries the long-term liability of glial scarring. Penetrating electrode arrays, including the Utah array that preceded Neuralink’s design, lose signal from more than 60% of electrodes within one year as scar tissue accumulates around the implant sites, according to published assessments.

Synchron’s Stentrode takes the opposite approach. It is delivered via catheter through the jugular vein and lodged in the superior sagittal sinus — the draining vein above the motor cortex — where it records motor intent signals through the vessel wall. No craniotomy, no cortical penetration. Signal bandwidth and spatial resolution are lower than intracortical systems, but the trade-off is a 20-minute median deployment time and a safety profile validated across ten patients with no device-related serious adverse events resulting in death or permanent increased disability over the 12-month evaluation periods. (SWITCH study, JAMA Neurology; COMMAND results)

Precision Neuroscience occupies a middle position. Its Layer 7 system is a thin-film subdural array — 1,024 electrodes on a flexible mesh one-fifth the thickness of a human hair — that rests on the cortical surface without penetrating tissue. It achieves higher resolution than the endovascular approach while avoiding the chronic scarring risk of penetrating electrodes. The current 510(k)-cleared version is limited to 30-day implant durations, positioning it initially as an intraoperative and short-term monitoring device rather than a chronic assistive technology. (GlobeNewswire — Precision FDA Clearance)

What the Clinical Data Actually Shows

Two of the three companies have peer-reviewed publications; one does not.

Synchron’s SWITCH study, published in JAMA Neurology in January 2023, remains the most rigorously documented result in the field. Over 12 months in four patients with ALS and other severe paralysis conditions, the Stentrode produced zero device-related serious adverse events and demonstrated that two participants achieved at least 92% accuracy in controlling text applications within three months of use. (Business Wire — Synchron JAMA Neurology publication)

The COMMAND study, completed in 2024 across six patients at three US centers, extended the safety dataset but has not yet appeared in a peer-reviewed journal. Results were presented at a 2024 conference: all six patients met the primary endpoint (no device-related serious adverse events over 12 months), and all demonstrated the ability to translate motor intent into digital outputs. (NeuroNews — COMMAND results) The absence of journal publication is a meaningful gap — efficacy details, quantitative digital task performance, and patient demographics remain confined to conference presentations.

Precision Neuroscience published in Nature Biomedical Engineering in October 2025. The paper combined preclinical animal studies with pilot data from five patients undergoing standard neurosurgical procedures. The Layer 7 system achieved high-resolution neural recordings sufficient for motor, sensory, and volitional decoding. The system generates 90 GB/h of electrophysiologic data — a data-management challenge that the Medtronic StealthStation integration is partly designed to address. (Nature Biomedical Engineering — Layer 7 paper)

Neuralink’s PRIME data remains almost entirely proprietary. The company publishes blog-format updates rather than peer-reviewed results. By February 2025, the five PRIME participants had collectively accumulated over 4,900 device-hours across more than 670 implant-days; by September 2025, 12 participants had logged more than 15,000 hours. (All Health Tech — Neuralink Update) Reported typing performance reached a mean of 13.7 WPM at 92% character accuracy, with some participants demonstrating up to 40 WPM on virtual keyboard tasks. These figures come from company communications, not independent peer review.

AI’s Role in the Decoding Stack

Neural signals are noisy, non-stationary, and patient-specific. In Neuralink’s system, raw decoding alone reportedly produces error rates of around 40% in communication tasks — too high for practical use. All three companies are integrating machine learning at the decoding layer, but Neuralink’s May 2025 integration of Grok (xAI’s large language model) made explicit what had been implicit: generative AI is now doing meaningful work in the signal-to-intent pipeline. (MIT Technology Review — Neuralink Grok integration)

By surfacing likely word completions and plausible conversational responses, the language model compensates for decoding ambiguity — reducing effective word error rate to under 10% in closed-vocabulary tests. This is architecturally important: it means BCI performance is increasingly a function of the AI stack, not just electrode count or signal fidelity. Synchron has made a similar move, describing its Cognitive AI division as training models that learn from brain data in real time. (MedTech Dive — Synchron $200M Series D)

The practical implication for evaluating these systems: raw electrode metrics (channel count, signal-to-noise ratio) are necessary but not sufficient. In principle, a lower-fidelity electrode paired with a well-trained decoder could outperform a higher-density array with a cruder model — though this trade-off has not yet been demonstrated in published head-to-head comparisons. That possibility means hardware specifications alone will not predict clinical performance.

Enterprise and Regulatory Signals

No implantable BCI has FDA premarket approval. The regulatory landscape has two live pathways and one emerging one.

Synchron holds the first FDA IDE for a permanently implanted BCI (COMMAND) and is using its $200 million Series D — closed November 2025, led by Double Point Ventures, with Bezos Expeditions and Khosla Ventures among existing investors — to fund a pivotal trial in 2026. (MedTech Dive — Synchron $200M) A successful pivotal trial would make Synchron the first company to file for PMA on an implantable BCI.

Precision Neuroscience took a different route: the 510(k) pathway, cleared in April 2025. This classification (Class II, substantial equivalence to a prior cortical electrode) allows commercialization now, within the 30-day implant constraint. (GlobeNewswire — Precision FDA Clearance) The Medtronic partnership announced 12 January 2026 is the clearest signal of the commercial strategy: rather than build a de novo surgical workflow, embed Layer 7 into the StealthStation system already present in most major neurosurgery operating rooms. As of that announcement, the Layer 7 system had been tested in over 68 patients across six US medical centers. (GlobeNewswire — Precision/Medtronic partnership)

Synchron’s Apple BCI HID integration (announced May 2025) is equally instructive about how the software stack is being built before PMA exists. Stentrode users — still clinical trial participants — can now natively control iPhones, iPads, and Apple Vision Pro through Apple’s new BCI Human Interface Device protocol. (MobiHealthNews — Synchron Apple integration) COMMAND participant Mark Jackson demonstrated navigating an iPad and composing text by motor intent alone. This is developer-ecosystem building, not a clinical outcome — but it determines what a physician can tell a patient the device will integrate with on day one of discharge.

Reimbursement is the unresolved variable. No CMS coverage decision exists for implantable BCIs. Until a PMA clears and Medicare establishes a coverage determination, hospital adoption outside clinical trials faces significant economic barriers. The global BCI market reached approximately $2.93 billion in 2024, with hospitals and neurology institutes holding the majority revenue share — but that figure is dominated by non-implantable EEG systems and research devices, not chronic implants. (Precedence Research — BCI Market)

What to Watch

• Synchron’s pivotal trial design, expected to be announced in 2026, will set the evidentiary standard for first-ever PMA approval of an implantable BCI. The choice of primary endpoints — functional communication tasks, activities of daily living, or a composite — will define what “efficacy” means for the field and what competitors must match.
• Peer-reviewed COMMAND data. The 12-month results presented in 2024 have not appeared in a journal. Publication — and the quantitative performance data it should contain — is the next meaningful disclosure from Synchron.
• Neuralink’s first peer-reviewed publication. The company has produced 15,000+ hours of device use data across 12 participants as of late 2025. None of it has appeared in a peer-reviewed journal. A Nature or New England Journal of Medicine publication would significantly change how clinicians and procurement committees assess the system.
• Precision Neuroscience’s path beyond 30 days. The current 510(k) clearance limits the Layer 7 system to 30-day implant durations. Whether Precision pursues a chronic implant pathway — and on what timeline — will determine whether it competes directly in the assistive technology market or remains primarily an intraoperative tool.
• CMS reimbursement pathway. The first company to clear PMA and secure a Medicare coverage determination will have a structural advantage: hospital systems will not build implantation programs for devices with no reimbursement code. The regulatory win and the billing code win are separate races, and the billing code may take longer.

Further Reading

• AI and Drug Discovery: What AlphaFold-Era Clinical Trials Actually Show
• EU AI Act Enforcement: What Enterprise Compliance Actually Requires in 2026

This article was produced with AI assistance and reviewed by the editorial team.