Source-audited records
261 source records back the entities, relationships, and claims. Gaps are listed, not filled.
Who holds authority over AI safety in China and Asia-Pacific, and what they do with it.
A stakeholder intelligence atlas for researchers, bridge-building organizations, and policy teams working from source-backed public evidence. Claims first, sources attached.
The authority stack against the safety spectrum, one node per organization. Every placement carries its reasoning.
Briefings
The AI+ campaign: who pushes, who pays, where 安全 fits
The August 2025 State Council opinion made AI adoption a national mandate. The briefing traces the instrument chain from the central plan through ministries, SOEs, and state finance to the provincial race it set off, separating sourced facts from marked judgments.
Read the briefingConservative claims
Confidence, sensitivity, and primary-source flags are visible in the public interface.
Relationship graph
Only public-safe relationship edges are shown, using conservative labels from the controlled taxonomy.
What's inside
The atlas in numbers
92
Entities loaded
92
Source-backed entities
4
Needs primary source
60
Public relationships
Last updated 2026-06-14. See the changelog.
Latest insight
Where the public record resists
Evidence thinness is not just a gap. It tells readers which claims the Atlas can settle, which it can only bracket, and where lab-level public records lag institutional records.
Read insightCountry slices
Static country pages
East Asia
China
China now has source-backed standards/evaluation records, dialogue/event records, frontier-developer disclosure records, and public technical-safety research records. These add PKU alignment tooling, BAAI FlagSafe/FlagEval, OpenCompass, BIGAI TongTest, and RUC interpretability/safety-alignment signals without treating generic AI work as frontier safety.
East Asia
Japan
Japan has a clear public-source technical safety stack through NII LLMC, LLM-jp Safety WG, Japan AISI guidance, and Japanese benchmark work from SB Intuitions and RIKEN-linked teams.
East Asia
South Korea
South Korea’s public technical-safety work is anchored by Korea AISI plus university evaluation work at Soongsil and method-level safety research at KAIST.
Southeast Asia
Singapore
Singapore’s public technical-safety record is strongest in open testing infrastructure and university-based methods: AI Verify Foundation / Moonshot, SMU agent-safety and model-security work, and NUS TrustLLM plus formal-methods framing.
Southeast Asia
Regional
ASEAN-wide profile for regional AI governance implementation. The register adds source-backed voluntary guidance, a generative-AI guide expansion, the Bangkok Digital Declaration, and a Malaysia official ASEAN AI Safe announcement while holding roadmap/search-snippet claims pending direct text.
Southeast Asia
Malaysia
Malaysia has an official AI roadmap, voluntary national AI governance and ethics guidance, a launched National AI Office, public-sector AI adoption guidance, and a documented but still-under-development National AI Action Plan 2026–2030 process.
Southeast Asia
Thailand
Thailand’s retrieved official materials show a layered governance posture: ETDA’s AI Governance Center, a 2019 AI ethics guideline, a 2023 executive AI governance guideline, a 2022–2027 national AI strategy page, and a 2025 consultation on draft AI-law principles.
Southeast Asia
Vietnam
Vietnam is the clearest ASEAN national case with binding instruments: official sources show a 2021 AI strategy, a dedicated AI Law enacted in December 2025, a February 2026 implementation plan, and an April 2026 implementing decree.
Southeast Asia
Indonesia
Indonesia’s strongest directly retrieved official evidence is the 2023 ministerial circular on AI ethics plus a BRIN page indicating an AI and cybersecurity research centre under OREI.
Southeast Asia
Philippines
The strongest official Philippine evidence captured here is implementation-oriented under DOST-ASTI: an AI project portfolio page, the February 2026 launch of NAICRI, and a year-one flagship-program article linking NAICRI to ACABAI-PH.
East Asia
Hong Kong
Hong Kong’s public technical-safety picture is thinner but source-backed: HKUST materials support LLM security work and HKGAI-V1 supports region-specific alignment and safety framing.