A man who had not moved his hands in years sat down, imagined closing his fist, and watched his own fingers curl shut. Months later he was pinching a glass bead between two fingers—and lifting a strawberry to his mouth. The bridge between his intention and his hand is a Chinese brain implant (脑机接口) barely the size of a coin.
The device behind the story
The system is BeiNao-1 (北脑一号), a semi-invasive brain-computer interface (脑机接口) built by the Beijing Institute for Brain Disorders together with a local neurotech firm, Xinzhida. "Semi-invasive" is the crucial word: surgeons open a small skull window and lay a flexible electrode array on the outside of the dura—the protective membrane around the brain—without piercing the tissue itself. That trades a little signal sharpness for a much safer, lower-infection profile than fully implanted electrodes.
The hardware is compact and wireless: a 128-channel array that records neural signals and beams them out without a tether, which matters because a wire through the skin is a lifelong infection risk.
What the trial actually shows
BeiNao-1 entered a multi-center registered clinical trial—the formal NMPA-regulated path toward a medical-device license—on 31 March 2026. The latest public figures, from Beijing municipal-science reporting and the institute itself, are:
- Nearly 30 human implants (the device maker reported 27 by mid-June 2026)
- Longest implant exceeding 15 months
- More than 70,000 cumulative safe operating hours across patients
- System performance described as stable throughout
The anchor patient is a person with a spinal-cord injury who was tetraplegic—all four limbs inert. Before the implant he could not move his hands at all. After surgery, imagining a fist let the system detect the motor signal, stimulate the right muscles, and close the hand. At the one-year follow-up he could grip coarse objects, pinch a glass bead with two fingers, and pick up a strawberry. A second line of work targets speech: the same platform is decoding familiar Chinese words for patients with amyotrophic lateral sclerosis (ALS), reaching close to a hundred common terms.
Why "semi-invasive" is the pragmatic bet
Fully invasive implants that sit inside the brain capture the richest signals but carry the highest surgical risk and the hardest long-term safety questions. BeiNao-1's epidural route is the compromise China's clinicians are betting on for real wards, not lab demos: enough signal to drive meaningful movement, low enough risk to implant in fragile patients.
The institute is also developing BeiNao-2 (北脑二号), a fully invasive 512-channel wireless system, for cases that need deeper resolution. Animal tests are done, and clinical validation is expected by the end of 2026. In monkey trials its decoding speed reached about 7.2 bits per second, close to the roughly 7.3 bits per second that Elon Musk's Neuralink has reported for its implant—a figure worth reading with care, since both numbers come from the companies themselves.
The timeline to a real product
This is still a trial, not a store shelf. The team aims to finish enrolling patients in 2026 and file for a Class III medical-device registration certificate in 2027. A Class III certificate is the highest-risk category in China's device regime; clearing it would let the system be sold as a therapy rather than used only under research consent.
Until then, every reclaiming of a strawberry is a data point from a small, carefully selected group—not a guarantee for the next patient who rolls into the ward.
Honest limitations
This article draws on People's Daily's national party-media platform, Beijing Municipal Science & Technology Commission reporting, and People's Daily Beijing, all covering BeiNao-1 and its lead researcher Luo Minmin. Keep these caveats in view:
- "World's highest level" and "global first" are the institute's own assessments, not independent benchmarks we could verify. The 128-channel, semi-invasive claim is strong on paper; head-to-head international comparison is not something this article can confirm.
- The trial is ongoing and small—roughly 27–30 implants. Anecdotes like the strawberry are powerful but are individual outcomes, not population statistics.
- The Neuralink comparison (7.2 vs 7.3 bits/sec) uses figures each company has reported; methodologies differ and we could not normalize them.
- The device is not yet approved for sale; 2027 is a planned filing date, not a granted license. Regulatory delay is possible.
- We relied on Chinese institutional and state-media sources; we did not access the trial's peer-reviewed publication or the full NMPA filing.
What readers can do now
- If you follow medtech, track BeiNao-1's NMPA Class III filing in 2027—approval, not the demo, is the moment this becomes a reimbursable therapy.
- If you invest in neurotech, separate "semi-invasive is safer" from "semi-invasive is proven"; watch enrollment completion in 2026 as the next hard milestone.
- If you care about the field's honesty, read single-patient recovery stories as signal, not proof—ask for the trial size and the peer-reviewed record before believing the revolution has shipped.