Link to the code: brain-emulation GitHub repository

Paradromics Connexus BCI: FDA Approval for Speech Restoration Trials 2026

Paradromics received FDA Investigational Device Exemption (IDE) approval for the Connect-One Early Feasibility Study in November 2025, becoming the first company cleared to test speech restoration using a fully implantable brain computer interface. The Connexus system enters human trials in Q1 2026 at three clinical sites across the United States, targeting patients with severe motor impairment who have lost the ability to speak.

The approval marks a significant development in the brain computer interface field, which has largely focused on cursor control and computer operation. Paradromics is pursuing direct speech restoration, a more complex decoding challenge that requires interpreting neural activity controlling the lips, tongue, and larynx in real time.

This is not speculative technology. Paradromics has demonstrated 200+ bits per second information transfer rates in preclinical models, over 20 times faster than the 4 to 10 bits per second reported by Neuralink’s first human patient. If replicated in human trials, this bandwidth could enable communication speeds approaching natural conversation.

Technology Readiness Level: TRL 3-4 (FDA approval for human trials, preclinical validation completed)

The Connexus System Architecture

The Connexus BCI consists of micro-electrode arrays, each electrode thinner than a human hair, that record neural activity from individual neurons in the motor cortex. The target region controls the articulatory muscles involved in speech production: lips, tongue, larynx, and jaw.

A single Connexus module contains 420 micro-needle electrodes. Multiple modules can be linked together, allowing the system to scale to over 1,600 recording channels. This electrode density exceeds most competing systems. Neuralink’s N1 implant features 1,024 electrodes across 64 flexible threads. Synchron’s Stentrode uses only 16 electrodes deployed through blood vessels.

The electrode materials combine titanium and platinum-iridium, chosen for biocompatibility and resistance to corrosion during long-term implantation. These materials have decades of documented safety in cardiac pacemakers and other implantable medical devices, reducing unknowns in the regulatory approval process.

Neural signals captured by the electrode array travel to a compact receiver implanted in the chest. This receiver wirelessly transmits data through the skin to an external computer. The computer runs AI decoding algorithms that translate patterns of neural firing into speech commands or computer control instructions.

This architecture differs from the wireless skull-mounted design used by Neuralink. Paradromics places the signal processing hardware in the chest rather than the skull, which may improve heat dissipation and simplify battery charging. The chest placement also reduces visible signs of the implant, which some patients may prefer for social reasons.

The system is fully implantable with no percutaneous connectors. All data and power transmission occurs wirelessly through the skin, reducing infection risk compared to older BCI designs that used wired connections penetrating the scalp.

Connect-One Early Feasibility Study Details

The Connect-One study will initially enroll two participants with impaired speech and limited extremity movement due to severe loss of voluntary motor control. Eligible conditions include amyotrophic lateral sclerosis (ALS), brainstem stroke, and high-level spinal cord injuries that affect both motor function and speech.

Clinical trial sites include:

  • UC Davis (Sacramento, CA) led by David Brandman, M.D., Ph.D.
  • Massachusetts General Hospital (Boston, MA) led by Daniel Rubin, M.D., Ph.D.
  • University of Michigan (Ann Arbor, MI) led by Matthew Willsey, M.D., Ph.D.

The trial protocol focuses on two primary outcomes: safety and speech restoration performance. Safety endpoints track adverse events including infection, bleeding, tissue damage, and device malfunction. Performance endpoints measure bits per second communication rate, word accuracy, and task completion time for speech and computer control tasks.

The study will evaluate long-term use of the Connexus BCI, though the specific duration has not been publicly disclosed. Early feasibility studies typically run 12 to 24 months, with potential extensions based on initial safety results.

Paradromics CEO Matt Angle stated, “This is the device that patients deserve.” Chief Medical Officer Stephen Ryu emphasized the trial would “demonstrate the unmatched performance, safety, and reliability of our high-bandwidth BCI to help overcome human limitations.”

Speech Restoration vs. Cursor Control

Most brain computer interface trials prioritize cursor control and computer operation. Patients learn to move an on-screen cursor by imagining hand movements, allowing them to type, browse the internet, and control devices. This approach has achieved measurable success across multiple research groups and companies.

Speech restoration presents a fundamentally different decoding problem. Natural speech production involves coordinating dozens of muscles in the face, throat, and respiratory system at millisecond timescales. The motor cortex generates complex, overlapping neural patterns that encode intended phonemes, words, and prosody simultaneously.

Decoding these patterns requires higher bandwidth than cursor control. A person speaking at normal conversational speed produces approximately 140 to 180 words per minute, equivalent to 200 to 300 bits per second when accounting for phonetic complexity and error correction requirements.

Paradromics claims its system achieves 200+ bits per second in preclinical models. If this performance translates to human patients, the Connexus BCI could support communication speeds comparable to natural speech. This would represent a qualitative improvement over current BCI systems that enable typing at 10 to 20 words per minute.

The company’s focus on speech restoration targets a specific patient population: people with locked-in syndrome, advanced ALS, or severe brainstem damage who retain cognitive function but have lost voluntary motor control. These patients often rely on eye-tracking devices or slow letter-by-letter communication methods. A high-bandwidth speech BCI could restore near-normal communication capabilities.

Comparison to Other BCI Approaches

The brain computer interface field includes several companies and research groups pursuing different technical strategies with varying risk-benefit profiles.

Neuralink has implanted 21 patients as of January 2026 with the N1 system, which features 1,024 electrodes on flexible polymer threads inserted into the motor cortex. Neuralink’s trials focus on cursor control and device operation, with speech restoration planned for future studies. The first patient achieved 4 to 10 bits per second communication rates, far below Paradromics’ claimed performance but sufficient for practical computer use.

Neuralink’s advantage lies in surgical automation. The company developed a precision robot that performs nearly the entire implantation procedure, reducing surgeon variability and enabling rapid scaling. Paradromics has not disclosed details of its surgical approach, which may limit deployment speed if manual neurosurgery is required.

Synchron uses an endovascular approach, deploying the Stentrode device through blood vessels to reach the motor cortex without open brain surgery. This minimizes surgical risk but provides lower spatial resolution with only 16 electrodes. Synchron has implanted several patients in the US and Australia, enabling cursor control and basic device operation. The company has not reported bits per second performance metrics comparable to Neuralink or Paradromics.

Precision Neuroscience received FDA 510(k) clearance for the Layer 7 cortical interface in April 2025. This system uses thin-film electrode arrays placed on the brain surface rather than penetrating the cortex. The approach may reduce tissue damage but provides lower signal quality compared to penetrating electrodes. Precision Neuroscience’s system is designed for temporary 30-day implantation periods rather than permanent placement.

Blackrock Neurotech has decades of experience with the Utah Array, a rigid silicon-based electrode grid used in numerous academic BCI studies. The Utah Array enables high-performance applications including robotic arm control and communication at 90 characters per minute in research settings. However, the rigid structure may cause more chronic tissue damage compared to flexible electrodes like those used by Neuralink and Paradromics.

Paradromics positions the Connexus system as a high-bandwidth solution specifically optimized for speech restoration. The electrode count exceeds Neuralink’s design (1,600+ vs 1,024), and the claimed 200+ bits per second bandwidth significantly surpasses all competing systems with published human data. Whether these advantages translate to superior clinical outcomes will depend on the Connect-One trial results.

Regulatory Pathway and Breakthrough Device Designation

Paradromics previously received FDA Breakthrough Device Designation, an expedited review pathway for medical devices that treat life-threatening or irreversibly debilitating conditions. This designation accelerates the development and regulatory approval process, allowing more frequent consultation with FDA reviewers and streamlined clinical trial requirements.

The IDE approval for Connect-One represents the next critical milestone. An IDE permits investigational devices to be used in human clinical studies to collect safety and effectiveness data. FDA grants IDE approval only after reviewing extensive preclinical data including animal studies, biocompatibility testing, and engineering validation.

Paradromics stated it is “the first company to receive IDE approval for speech restoration with a fully implantable BCI.” This claim distinguishes Connexus from partial or externally wired systems that have been tested for speech applications in academic settings.

The regulatory pathway for BCIs remains complex and evolving. The FDA classifies neural implants as Class III medical devices, the highest risk category, requiring premarket approval (PMA) before commercial sale. Early feasibility studies like Connect-One provide initial human data to inform larger pivotal trials that will ultimately support PMA applications.

Neuralink followed a similar pathway, receiving FDA approval for the PRIME Study in May 2023 after initial rejection and protocol refinement. The FDA’s experience reviewing Neuralink’s application likely informed its evaluation of Paradromics’ submission, potentially accelerating the approval timeline.

Technology Pipeline Beyond Speech Restoration

Paradromics is developing additional BCI applications beyond the Connect-One study. The company mentions Tempo, a brain computer interface designed for mental health applications including chronic pain, addiction, and depression treatment.

Tempo targets brain state monitoring rather than motor control or speech. The system would detect neural signatures associated with chronic pain episodes, addictive cravings, or depressive states, potentially enabling closed-loop interventions such as targeted brain stimulation or neurofeedback therapy.

This application represents a different use case for high-bandwidth neural recording. Mental health BCIs require decoding distributed neural activity across multiple brain regions rather than focusing on the motor cortex. Whether the Connexus electrode architecture can be adapted for this purpose remains unclear based on publicly available information.

The chronic pain market alone represents a massive potential application. Over 20% of US adults experience chronic pain, and existing pharmaceutical and surgical treatments have significant limitations. A BCI that could detect and modulate pain-related neural activity could address a major unmet medical need, though clinical validation for such applications is likely years away.

The Speech Bandwidth Challenge

Achieving 200+ bits per second bandwidth in human patients will require solving several technical challenges that extend beyond electrode hardware.

Neural decoding algorithms must map high-dimensional neural activity patterns to intended speech sounds in real time. The motor cortex does not encode phonemes directly. It generates continuous control signals for dozens of articulatory muscles. Machine learning models must learn the mapping between neural firing patterns and acoustic output for each patient individually.

Calibration and training present practical obstacles. Patients will need to complete training sessions where they attempt to speak or imagine speaking specific words and sentences. The BCI system records neural patterns during these attempts and trains decoding models to recognize the patient’s unique neural signatures. If the patient cannot produce any vocalization due to complete paralysis, calibration may require imagined speech only, which typically yields lower accuracy.

Signal stability over time affects long-term performance. Neural recordings from implanted electrodes can degrade due to scar tissue formation, electrode shift, or changes in neural firing patterns. Paradromics claims the biocompatible materials reduce chronic inflammation, but multi-year stability data does not yet exist from human implants.

Error correction and feedback will be necessary for practical communication. Even with 200+ bits per second bandwidth, decoding errors are inevitable. The system must provide real-time feedback so users can detect and correct mistakes, similar to how able-bodied speakers self-monitor and adjust their speech.

Academic research groups have demonstrated speech BCI prototypes achieving word error rates of 5 to 10% in patients with intact speech. Extending these results to patients with severe motor impairment who cannot produce any vocalization represents a significant challenge that Connect-One aims to address.

Timeline and Path Forward

The Connect-One Early Feasibility Study launches in Q1 2026 with two initial participants. Successful completion of this trial could lead to expanded enrollment and eventually pivotal trials required for FDA approval.

The timeline from early feasibility studies to commercial approval typically spans 5 to 10 years for Class III medical devices. Neuralink began its first human implant in January 2024 and remains years away from commercial availability. Paradromics should expect a similar timeline unless breakthrough clinical results accelerate the regulatory process.

The company raised $33 million in previous funding rounds and received Breakthrough Device Designation, indicating financial and regulatory support for continued development. However, BCI development requires sustained investment over many years. The company will likely need additional funding to complete clinical trials and scale manufacturing.

Paradromics’ stated goal is to “help overcome human limitations,” a broader vision than speech restoration alone. The Connexus platform’s high electrode count and bandwidth could support future applications including sensory restoration, cognitive enhancement, or brain-to-brain communication, though these remain speculative.

For now, the focus remains on the immediate clinical challenge: demonstrating safe, effective speech restoration in two patients with severe motor impairment. If Connect-One succeeds, Paradromics will have validated a new class of medical device that could restore natural communication capabilities to thousands of people living with locked-in syndrome, advanced ALS, and other devastating neurological conditions.

The speech restoration problem is harder than cursor control. The potential impact is also greater. Natural conversation remains one of the most fundamental human capabilities. A BCI that can restore it would represent a genuine breakthrough in neurotechnology, not just an incremental improvement over existing systems.

Official Sources

Primary Sources:

Comparative BCI Research:

Technical Reviews:

Clinical Trial Information:

  • Paradromics Clinical Sites: UC Davis (David Brandman, M.D., Ph.D.), Massachusetts General Hospital (Daniel Rubin, M.D., Ph.D.), University of Michigan (Matthew Willsey, M.D., Ph.D.)
  • FDA Breakthrough Device Designation and IDE Approval (November 2025)

Related Articles: