NHSIA Releases 9th Batch of Intelligent Oversight 'Two Libraries'

On April 26, 2026, China’s National Healthcare Security Administration (NHSA) officially released the ninth batch of its intelligent supervision ‘two libraries’—comprising a knowledge base and a rules base. This update marks a pivotal step in aligning domestic AI-driven medical device regulation with international clinical evidence expectations, directly influencing export readiness for remote surgery systems and AI-assisted diagnostic hardware across key global markets.

Event Overview

The NHSA launched the ninth batch of its intelligent supervision ‘two libraries’ on April 26, 2026. For the first time, the updated libraries explicitly incorporate three high-complexity clinical-AI scenarios: intraoperative brain–computer interface (BCI) monitoring, real-time path verification for surgical robots, and multimodal imaging AI-assisted diagnosis. The framework is designed to support dynamic, risk-based oversight of algorithmic behavior, data provenance, and clinical workflow integration—not as static certification criteria, but as an evolving reference for regulatory alignment.

Industries Affected

Direct Export Enterprises: These firms face revised expectations from overseas importers and regulators—particularly under the EU Medical Device Regulation (MDR), Saudi Food and Drug Authority (SFDA), and emerging ASEAN frameworks. The ‘two libraries’ do not constitute foreign regulatory requirements, but are increasingly cited by international partners as a de facto benchmark for evaluating clinical data traceability, algorithm version control, and closed-loop validation rigor. Export documentation packages may now need to map internal clinical evidence against specific ‘two library’ logic nodes (e.g., how robot path deviation thresholds correspond to defined safety rules).

Raw Material & Component Suppliers: Firms supplying certified sensors, neural signal processors, or validated imaging detectors will see heightened demand for audit-ready technical dossiers—including test reports tied to clinical use cases referenced in the new libraries (e.g., BCI latency under simulated intraoperative interference). Supply chain transparency—especially regarding firmware version history and calibration metadata—becomes a differentiator, not just compliance overhead.

Contract Manufacturing & OEM Producers: Manufacturers performing final assembly, software integration, or clinical validation services must now align their quality management system (QMS) records with the ‘two libraries’’ operational definitions—for example, documenting how AI inference latency was measured across multimodal imaging pipelines during verification testing. Their role shifts from component-level compliance to end-to-end clinical logic accountability.

Regulatory & Clinical Support Service Providers: CROs, regulatory consultants, and clinical evidence strategists are seeing increased client requests for ‘two library mapping’—a service that crosswalks existing clinical study protocols, algorithm validation reports, and post-market surveillance plans against relevant knowledge entries and rule triggers. This is not yet mandatory abroad, but functions as a strategic risk-mitigation layer for faster dossier acceptance.

Key Focus Areas and Recommended Actions

Map clinical validation scope to newly listed scenarios

Exporters should conduct a gap analysis between current clinical evidence packages and the three newly included use cases (BCI monitoring, surgical robot path verification, multimodal AI diagnosis). Where gaps exist, prioritize retrospective data reanalysis or targeted supplementary studies—not full revalidation—unless algorithm architecture changes are involved.

Strengthen algorithm version governance and traceability

The ‘two libraries’ emphasize auditable links between deployed AI versions, training data lineage, and clinical outcomes. Firms should formalize version control logs that include model hash values, input data provenance tags, and corresponding rule-library trigger conditions—structured to meet SFDA Annex 14 or EU MDR Annex XIII expectations.

Prepare modular regulatory dossiers aligned with ‘two library’ logic

Instead of monolithic submissions, structure technical documentation into reusable modules (e.g., ‘Path Verification Logic Module’, ‘Multimodal Fusion Validation Module’) that can be selectively referenced and updated per market—reducing duplication while improving responsiveness to evolving library iterations.

Editorial Perspective / Industry Observation

Observably, the NHSA is no longer treating AI regulation as a domestic containment tool—but as a strategic enabler of global interoperability. The ‘two libraries’ lack binding extraterritorial force, yet their granular, clinically grounded design gives them persuasive weight among regulators lacking mature AI-specific infrastructure. Analysis shows this reflects a broader shift: from ‘certification-first’ to ‘evidence-readiness-first’ export strategy. From an industry perspective, the real value lies not in conformity, but in using the libraries as a diagnostic lens—revealing where clinical data collection practices lag behind international expectations for algorithmic accountability.

Conclusion

This release does not create new legal obligations outside China, but it recalibrates the evidentiary floor for credible AI medical hardware exports. It signals that international market access increasingly hinges on demonstrable clinical logic coherence—not just technical performance. A rational conclusion is that firms treating the ‘two libraries’ as a living reference, rather than a static checklist, will gain measurable advantage in registration timelines and importer trust.

Source Attribution

Official announcement issued by the National Healthcare Security Administration (NHSA) on April 26, 2026. Full text available via www.nhsa.gov.cn. Note: Future library updates (e.g., inclusion of generative AI diagnostics or federated learning workflows) remain under observation; no timeline has been published for the tenth batch.