A child with attention difficulties trains focus with a headband. A wheelchair user steers by thought at a sports event. None of it required surgery or a hospital ward. China just gave that "beyond-the-hospital" brain-tech a common rulebook.
The move targets a specific gap. Invasive BCI — the kind that goes under the skull — has grabbed the headlines and the regulators. But the larger near-term market is non-invasive neurotech (非侵入式脑机接口) that helps people rehabilitate, focus, or move without ever entering an operating room. Until now, that market had no shared standard.
The three standards
Three documents frame the change:
- GB/T 47023-2026《信息技术 脑机接口 参考架构》 (reference architecture) and GB/T 47127-2026《信息技术 脑机接口 多模态数据格式》 (multimodal data format). Published January 28, 2026 via the SAMR announcement No. 4/2026, they took effect August 1, 2026. They are the first foundational national standards from the BCI subcommittee (全国信息技术标准化技术委员会脑机接口分会, TC28/SC43).
- GB/T《信息技术 非侵入式脑机接口康复训练系统技术要求》 (technical requirements for non-invasive BCI rehab training systems). Filed March 2, 2026 (SAMR announcement No. 15/2026), on an 18-month cycle, led by Pazhou Lab (琶洲实验室) with professors Li Yuanqing (李远清) and Pan Jiahui (潘家辉). It applies to the design and quality assessment of rehab instruments and is explicitly scoped as "not a medical device" (不涉及医疗器械产品).
The "data dialect" problem
Before these rules, BCI data had no common shape. One lab recorded with 64 channels, another with 128 or 256. Sampling rates ranged from as low as 2 Hz to as high as 1,000 Hz. Labeling rules differed lab to lab. The result, in the standard-setters' own words, was data that behaved "like files written in different dialects" — impossible to compare or reuse.
The data-format standard (47127) sets storage structure, exchange protocols, and encoding for EEG, eye-tracking, and near-infrared signals. In plain terms, it lets a rehab device from one vendor and a training dataset from another actually talk to each other — which is what large-scale efficacy evidence and shared algorithms require.
Why rehab, and why "beyond the hospital"
The demand is concrete. China reports more than 24.72 million people with physical disabilities (肢体残疾) and over 100 million with mental disorders (精神障碍). Non-invasive BCI has shown promise in rehab assistance, consciousness assessment, and awakening therapy — but a lack of standards had been throttling industrialization.
Pazhou Lab already illustrates the beyond-hospital path. It has put non-invasive EEG-control products into real use: a BCI AI wheelchair (脑机AI轮椅) that users steer by thought, demonstrated at the National Games for persons with disabilities (残特奥会); an attention-training system (专注力训练系统) aimed at children with ADHD and autism; and a mindfulness system (正念冥想系统) used in competitive training and education. The lab says it has five such products on the market.
What the standard does NOT do
- These are recommended (GB/T) standards, not mandatory law. They set a quality and interoperability baseline, not a market ban.
- The rehab-specific standard is still being drafted (filed, not yet published); the 18-month clock started in March 2026.
- They deliberately do not cover invasive medical devices — that is the National Medical Products Administration's lane (separate from this article's subject).
Honest limitations
- GB/T standards are recommended, not legally binding; their bite depends on whether procurers and regulators adopt them.
- The rehab standard is in drafting, not in force; we could not confirm its final technical thresholds.
- The "five products" and "24.72 million disabled / 100 million mental-disorder" figures come from Pazhou Lab and state-media reporting, not an independent audit.
- A standard solves interoperability and quality certification, not clinical efficacy — a certified rehab device can still be ineffective.
- The line between "medical device" and "rehab instrument" here is a regulatory boundary, not a technical one, and will shift as rules evolve.
What readers can do now
- If you build non-invasive BCI, align your data pipeline to GB/T 47127 now — interoperability is becoming the procurement question, not a nice-to-have.
- If you buy rehab tech, ask vendors which products are designed to the new rehab GB/T; early alignment signals a serious supplier.
- If you follow policy, mark the ~18-month window from March 2026 and watch for the rehab standard's publication — that is when "beyond the hospital" neurotech gets its first real quality ruler.