Organizations
Full dossiers exist for organizations where the source record is deep enough to support one: a claim about why the organization matters, at least two assessed importance aspects, and three or more source records. Organizations below that bar keep their card in the atlas.
国家互联网信息办公室
Every public-facing generative AI service in China clears rules CAC issued or co-issued; its filing, security-assessment, and labeling architecture is the enforcement layer the rest of the stack builds on.
工业和信息化部
The industry ministry whose name is on every joint AI rule: co-issuer of the binding service rules and the 2024 standardization guide, and home ministry of CAICT and CESI, the institutes that run much of China's AI evaluation.
全国网络安全标准化技术委员会
The committee where China's AI security expectations become testable documents: its guides pair with binding rules, its framework names the risk categories, and its new WG9 is where AI security standards are being drafted.
人工智能安全标准工作组
The working group TC260 created to draft China's AI security standards; a frontier-lab scientist leads it, making it the most direct lab-to-standards channel on public record.
中国信息通信研究院
The MIIT institute that built much of China's AI evaluation machinery: the AIIA alliance, the AI Safety Benchmark, and the Fangsheng benchmark system all run through it.
中国电子技术标准化研究院
The standards institute where China's formal AI standards get drafted: secretariat unit for the national AI subcommittee, publisher of a general large-model evaluation standard, and builder of the first AI national-standard benchmark system.
中国人工智能产业发展联盟
The industry alliance through which Chinese AI firms sign safety commitments and get benchmarked; CAICT runs its machinery, which makes it the main soft-governance channel between the state institutes and the model developers.
OpenCompass 司南
Shanghai AI Laboratory's open evaluation system is one of the few public windows into how a Chinese frontier lab measures model capability, and it carries safety as an explicit dimension.
清华大学战略与安全研究中心
The Tsinghua center that co-hosts the recorded U.S.-China and China-EU AI security dialogues; when track-2 talks on AI risk happen, this is the counterparty the public record shows on the Chinese side.