A paralysed patient in a Shanghai ward moves a robotic hand by thinking about it. A researcher in Beijing watches a screen where a silently imagined sentence turns into typed Chinese. In hospital logs across the country, nurses are recording another day without a serious adverse event.
In March 2026, China's regulator did something no major health authority had done before: it approved a brain–computer interface (脑机接口) for routine clinical use. That certificate grabbed the headlines. The more interesting story is what has happened since — a quiet clinical wave in which at least three more Chinese teams have pushed implants into human bodies, not as science demos but as registered studies.
The approval that reset the clock
The device is NEO, built by Neuracle Medical together with Tsinghua University. On 13 March 2026 the National Medical Products Administration cleared it as the world's first approved invasive (technically semi-invasive, epidural) BCI Class III medical device. Within days the system received a medical-insurance coding in Shanghai, and the first implant surgery after approval was completed at Huashan Hospital in July 2026.
NEO is deliberately modest in its ambition: a coin-sized eight-electrode array sits on top of the dura mater, outside the brain tissue, and translates the intent to grasp into movement of an external pneumatic glove. It does not cure spinal-cord injury. It routes around it. That trade — less invasive, narrower signal — is exactly why a regulator was willing to clear it first.
The wave behind the certificate
Approval is a starting line, not a finish. Three parallel Chinese programmes show how fast the field is now moving.
Beinao (北脑): from 128 channels to 512
The Beijing Brain Institute (CIBR) and partner Xinzhida NeuroTech built Beinao-1, a 128-channel wireless fully-implanted semi-invasive system. By April 2026 it had been implanted in roughly 10 patients, with the device logging more than 45,000 cumulative safe in-body hours and no reported serious adverse events. Its upgrade, Beinao-2, is a 512-channel system that has completed animal trials and reached a brain-control information rate of about 7.18 bits per second; the institute expects it to begin human clinical validation within 2026.
NeuroXess (脑虎): the subdural challenger
Shanghai's NeuroXess took a different route — a flexible electrode placed under the dura, on the cortex surface, with the battery and wireless units split off under the chest skin to keep heat away from the brain. On 7 July 2026 it opened a GCP registrational clinical trial (NCT07720882) at Huashan Hospital under Professor Mao Ying, spanning 15 tertiary hospitals with a planned 32 patients who have cervical spinal-cord injury. The primary endpoint is expected in January 2027. On 21 July 2026 the system entered the NMPA's innovative-device "green channel" — the first domestic subdural flexible BCI to do so. NeuroXess cites full-link latency under 50 milliseconds and cursor decoding at 5.2 bits per second, and is building a 14,300-square-metre "Super Factory" in Jiangxi for second-half 2026 production at roughly 10,000-unit scale.
The rest of the field
Other teams are close behind. Stairmed (阶梯医疗) and Zhiruan Medical (智冉医疗) have moved invasive implants into early clinical or registrational stages. Capital is following: industry tracker Pandaily counted 64 BCI funding rounds in China in the first half of 2026 worth 7.25 billion yuan (≈ US$1.0B / HK$7.9B) — more than the full-year 2025 total.
"Approved" is not "proven"
The speed is real, but so are the caveats.
- Sample sizes are small. Ten implants and 32 planned enrolments are early-phase numbers, not population evidence.
- Long-term stability is unknown. Electrode signal quality over years, infection risk, and device lifespan need five-to-ten-year follow-up.
- Routes are not interchangeable. NEO's epidural design is safer but lower-resolution than Neuralink's 1,024-channel cortical threads. China's "first" is regulatory, not a claim of superior signal.
- Reimbursement is uneven. Insurance coverage exists for some products in some cities — not for everything marketed under the "BCI" label.
What readers can do now
- If you or a family member are affected, ask "is it suitable" before "is it advanced." BCI has strict indications; a qualified centre and a full file beat any online article.
- Check two things: approval number and reimbursement. One without the other means a device you cannot legally or affordably get.
- Treat consumer "brain headsets" as wellness gadgets, not medicine. Several-thousand-yuan trainers are not the same as a registered brain–computer interface (脑机接口).
Honest limitations
This article draws on Chinese state media (Xinhua, China National Radio, Science and Technology Daily), official institute and company disclosures, and English trade coverage (Pandaily). Most patient counts, latency and decoding-rate figures come from the developers or their clinical centres, not from independent audits. Trial enrolment numbers are targets or early reads, not completed outcomes. "World's first approved invasive BCI" refers specifically to NMPA's Class III clearance of a semi-invasive device; it is not a statement about cortical penetration depth or signal quality versus competitors. Where a figure had only a single source (e.g. the H1 2026 funding total), it is attributed as such. Reporting is current to late September 2026; BCI timelines are moving quickly.