Country profile · China
China: sovereignty, control and biomedical AI capability
China linked digital sovereignty and controllable AI with beneficial applications and wider participation in governance. BioMedAgent is a published computational research example involving Chinese Academy of Sciences institutes, universities and hospitals; its benchmark and workflow results do not establish patient benefit.
Fu Cong
Permanent Representative of China to the United Nations
UN Security Council meeting 10228 · 23 September 2026
UN position
What China argued
UN proposition — paraphraseSecure, controllable AI should combine digital sovereignty, beneficial uses and UN-centred cooperation.
China’s intervention argued that countries should retain the ability to choose AI products and services according to their needs. It criticised exclusive technology groupings and pressure to take sides, presenting digital sovereignty and open cooperation as compatible with shared scientific and technological progress. These are China’s policy arguments, not independently measured effects of technology restrictions.
The speech also called for stronger regulation, emergency response and cross-border cooperation on misuse. Companies should accept legal regulation and public scrutiny. In military uses, the intervention stressed restraint, civilian protection and human authority over ultimate decisions, alongside international law and the UN Charter.
Public health was among the beneficial uses named, and the intervention called for a stronger voice for developing countries in AI governance through the United Nations. BioMedAgent provides an example of biomedical research capability; its existence does not validate China’s political claims about sovereignty, technology blocs or governance.
Key themes
Digital sovereignty and national choiceSecure and controllable AICompany responsibility and public scrutinyHuman authority in military decisionsBeneficial applications including public healthUN-centred cooperation and participation of developing countries
Scientific context
Selected AI × science example
Computational/methods researchBioMedAgent is a published computational methods research system for assembling multi-agent biomedical analysis workflows with specialised tools. The research involves Chinese Academy of Sciences institutes, universities and hospitals. The paper reports benchmark evaluation and biomedical analysis tasks, including analysis across biological datasets and pathology segmentation. These workflow and benchmark results do not establish clinical validation, improved patient outcomes or patient benefit.
Evidence boundary
What the evidence does show
The BioMedAgent paper reports a multi-agent workflow for biomedical analysis using specialised tools, with benchmark evaluation and selected research tasks. The collaboration includes Chinese Academy of Sciences institutes, universities and hospitals.
What it does not show
Benchmark or workflow performance does not establish clinical validation, safe use in clinical decisions or improved patient outcomes. BioMedAgent is not a clinically validated system, and patient benefit is not established. The selected research does not validate political claims about sovereignty, technology blocs or governance.
Speech–science connection
The intervention explicitly identifies public health and scientific advancement as potential benefits. BioMedAgent illustrates biomedical analysis capability, but does not establish health benefit or validate the intervention’s political claims about sovereignty, technology blocs or governance.
Sources
UN Security Council meeting 10228 · 23 September 2026. Diplomatic position is paraphrased; audio verification is incomplete.
China
Permanent Mission of China — Fu Cong meeting-specific remarks · Permanent Mission of China
Supports: speaker identity and meeting-specific intervention.
View sourceBioMedAgent · Nature Biomedical Engineering
Bu et al., Nature Biomedical Engineering, online 30 March 2026. DOI 10.1038/s41551-026-01634-6.
Supports: selected China-associated biomedical computational/methods example.
Evidence boundary: benchmark/workflow performance does not establish patient benefit.
View sourceDOI: 10.1038/s41551-026-01634-6