China Issues the World's First AI-Brain Interface Product Standard

Beijing Writes the Rulebook for the Data Behind Brain Implants

China's National Medical Products Administration on Sept 14 approved the release of the country's third brain-computer interface medical device standard, and what the regulator describes as the first product standard in the world for BCI devices that use artificial intelligence to process brain-signal data. It takes effect on Sept 1, 2027.

The document is narrower than its title suggests and more consequential than its length. It sets technical requirements and test methods for the entire data chain a modern BCI device depends on, from collection and processing to labelling, storage and access. Its stated purpose is to close a gap that has slowed commercialisation across the industry: the absence of a common yardstick for the quality and construction of the datasets used to train clinical algorithms.

For readers in Tokyo, the significance lies in the sequencing. Japan has staked its neurotechnology ambitions on a 2050 horizon. China is building the same industry with dated milestones, and it is using standards, not subsidies alone, to do it.

A Standard for Machines That Learn From Brainwaves

The NMPA's own explanation is deliberately plain. A BCI medical device, it said, works like a machine that reads brain signals, and it can only understand a patient's intention accurately after learning large volumes of electroencephalogram data and training an artificial intelligence algorithm on them. Those words matter because they describe the regulatory risk: if the training data is inconsistent, the decoder learns the wrong thing.

The new standard is a recommended industry standard rather than a mandatory one, which is normal practice in China's device framework. Recommended standards still carry weight in registration review, and the NMPA has been explicit that algorithmic error is a clinical risk. Under its June 30 classification guidance, a non-invasive device for stroke rehabilitation qualifies for the lighter Class II pathway only if it does not use AI-based intent decoding. Add AI, and the product escalates to Class III, the highest risk tier, because a misinterpreted intention can cause injury.

It also fits a pattern. The NMPA published a terminology standard for BCI medical devices, YY/T 1987-2025, on Sept 16, 2025, effective Jan 1, 2026. A second standard, covering sensing and response test methods for closed-loop implantable neural stimulators, has also been published. A national classification standard for BCI medical devices is in the pipeline. This week's release is the third in that series, and the first to reach into the data layer where AI actually decides what a patient's brain meant.

From Approval to Insurance Code in One Year

The standard did not arrive in a vacuum. In July 2025, seven ministries and agencies - the industry and information technology ministry, the national development and reform commission, the education ministry, the national health commission, the state asset regulator, the Chinese Academy of Sciences and the NMPA - issued a joint implementation plan for the BCI industry, numbered MIIT document 164 of 2025, dated July 23 and published on July 30.

That plan set two dated targets. By 2027: breakthroughs in core technology, a preliminary technology, industrial and standards system, electrodes, chips and finished devices at internationally advanced levels, and two to three industrial clusters. By 2030: a reliable industrial system, two to three globally influential leading companies, and overall strength among the world's front ranks. It also instructed ministries to build out technical, product, service and testing standards, to take part in international standards work, and to establish a data governance framework that prevents leaks of what it called brain privacy.

The clinical side has moved at a comparable pace. On March 13, 2026, the NMPA cleared registration of the NEO implantable hand motor function compensation system, developed by Shanghai-based Neuracle Technology with Tsinghua University, for patients with tetraplegia caused by cervical spinal cord injuries. The system places electrodes between the skull and the dura mater, uses wireless power and communication, and drives a pneumatic glove through decoded neural signals. Eligible patients are aged 18 to 60, with injuries between C2 and C6, stable for at least six months. CGTN reported that the National Healthcare Security Administration completed medical insurance coding for the device within a week, and by July 3 it reported that certain BCI procedures were covered by the health system. On July 16, Huashan Hospital, affiliated with Fudan University in Shanghai, performed China's first commercial BCI surgery using the system.

Dataset Quality Becomes the Industry's Gate

That speed is why the data standard matters more than it might appear. Every one of those approvals rests on evidence generated from Chinese patients, in Chinese hospitals, with algorithms trained on Chinese-language neural data - exactly what the seven-ministry plan called for, alongside novel electrodes, dedicated chips and signal decoding algorithms.

The commercial logic is straightforward. CCID Consulting estimated China's BCI market at 3.2 billion yuan in 2024, up 18.8 per cent, and projects 6.14 billion yuan by 2028, a compound annual growth rate of 17.7 per cent. The China Academy of Information and Communications Technology counted more than 1,000 publicly disclosed BCI financing deals worldwide and more than 300 funded companies as of the end of March 2025. Domestic deal flow is smaller but concentrated: 30 disclosed transactions worth 1.34 billion yuan between 2022 and 2024, with non-invasive developers taking 780 million yuan and invasive developers 560 million yuan. Companies including Huichuang Medical and NeuroXess have each raised three times.

China's BCI industry is at a critical stage characterized by accelerated technological breakthroughs, rapid ecosystem development and a gradual increase in commercialization, said Ming Dong, vice president of Tianjin University and a member of the national committee of the Chinese People's Political Consultative Conference. The country, he said, has already entered the first tier of global BCI development, with non-invasive technology at parity with leading international standards and the gap in invasive key components narrowing quickly.

China Moves First on Invasive Interfaces

The split between invasive and non-invasive devices is where China's regulatory choices become strategic. Its classification guidance sends every implantable BCI device to Class III under code 12-00, reflecting surgical risk and clinical uncertainty, while non-invasive devices are Class III by default with a single narrow exception. Implantable BCI devices also appear in the NMPA's priority review catalogue for high-end medical devices, and the agency runs a pre-submission model for innovative applicants built on early involvement, company-specific guidance and rolling review. That was the route used for the NEO approval.

In parallel, the national standards regulator issued a batch of 18 national standards in April 2026 covering intelligent connected vehicles, semiconductor devices, brain-computer interfaces and BeiDou navigation chips. BCI was named for the first time in the government work report as one of the country's industries of the future, and the 15th Five-Year Plan treats it as a priority future industry.

The result is a market where the rules arrive before the products scale: classification, nomenclature, terminology, and now the data that feeds the algorithms. In a field where reimbursement, liability and evidentiary standards all hang on data provenance, that is a deliberate order of operations.

Japan's Cybernetic Avatar Bet Runs to 2050

Japan has chosen a different emphasis, and the contrast is instructive. Its flagship neurotechnology effort sits inside the Moonshot Research and Development programme, whose Goal 1 - approved by the Council for Science, Technology and Innovation on Jan 23, 2020 - commits the country to a society free from the limitations of body, brain, space and time by 2050.

The vehicle is the cybernetic avatar: remote robotic or virtual proxies that extend a person's physical, cognitive and perceptual reach. The programme defines three classes, including a BMI-cybernetic avatar that uses brain-machine interfaces to restore communication for people with disabilities, and an in-body avatar for internal monitoring and treatment. Its nearer-term target is ambitious in a different way: by 2030, one person operating more than ten avatars on a single task at the same speed and accuracy as one.

Execution is spread across universities and corporate labs. The ATR-led project on intention-driven avatars, known as the Internet of Brains, brings together Keio University, the University of Tokyo, Sony Computer Science Laboratories, the University of Osaka, the National Institute for Physiological Sciences and the University of Melbourne. At Osaka, neurosurgeon Takufumi Yanagisawa works on electrocorticography-based interfaces with epilepsy patients, on restoring communication for people with amyotrophic lateral sclerosis, and on minimally invasive endovascular electrodes threaded into cerebral blood vessels - an approach he has described in direct comparison with the stent-electrode technology being developed by the American firm Synchron. In a project survey of ALS patients, more than half said they would want an invasive brain-machine interface, with communication support the priority.

Japan's strengths are real and upstream. Brain/MINDS has mapped primate brains since 2014, the DecNef method for decoding and modifying brain activity emerged from Japanese laboratories, and the country led whole-brain simulation work on the K computer. What Japan has not built in parallel is the regulatory and reimbursement machinery that turns that science into billable clinical practice. Device oversight runs through the Pharmaceuticals and Medical Devices Act and the PMDA's existing framework, while the Moonshot programme looks past the clinic to a 2050 social vision.

That is a defensible bet on basic science. It also means China, which started later in several underlying technologies, is now the country defining what a compliant BCI device looks like, how its data must be assembled, and how quickly a patient can be reimbursed for one.

What to Watch For

Three things. First, the national classification standard now in the standards pipeline will determine how the NMPA's June guidance hardens into binding category rules - the point at which foreign developers must decide whether to register in China or stay out. Second, watch how the new data standard is applied in review: a recommended standard becomes a practical gate when reviewers begin asking for dataset documentation, and developers holding non-standard or non-Chinese datasets will feel it first.

Third, watch whether Japan's data governance keeps pace with its hardware ambitions. Tokyo has been an active voice in international AI and neurotechnology ethics discussions, and the Moonshot projects carry their own ethics and legal workstreams, including dedicated research on neurolaw. But standards are how industries are shaped in practice. In BCI, the sequence of terminology standard, classification guidance and now AI data requirements gives China a head start in writing a template other regulators will have to answer.

For Japanese suppliers of electrodes, chips and surgical robotics, the near-term opportunity is real. The seven-ministry plan explicitly calls for high-channel-count acquisition chips, ultra-low-power processing chips and sub-micron precision implantation robots. The longer-term question is whether Japan participates in the standard-setting or reacts to it.

By Kenji Tanaka, Staff Writer

This article was produced with AI-assisted research and editorial support. Sources: CGTN, CCTV News, NMPA, People's Daily Online, MIIT, CCID Consulting, JST Moonshot R&D.