[
  {
    "id": "cn-cac",
    "name_en": "Cyberspace Administration of China",
    "name_local": "国家互联网信息办公室",
    "acronym": "CAC",
    "entity_type": "government",
    "country": "China",
    "city": "Beijing",
    "region": "East Asia",
    "summary": "Core Chinese internet-governance regulator for AI-related online information services. Primary-source materials show CAC co-issuing the 2023 Interim Measures for Generative AI Services, the 2025 AI-generated content labeling measures, the 2022 deep-synthesis rules, and the 2024 national AI standardization guide.",
    "focus_areas": [
      "ai_governance",
      "policy_regulation",
      "content_safety",
      "standards"
    ],
    "tags": [
      "regulator",
      "content governance",
      "co-issuer",
      "AI service rules",
      "generative AI service governance",
      "deep synthesis governance",
      "algorithmic recommendation governance",
      "AI content labeling",
      "security assessment for certain internet AI services",
      "AI standards policy coordination",
      "China"
    ],
    "source_ids": [
      "report-china-standards-evaluation-2026-06-05",
      "cac-generative-ai-measures-2023",
      "cac-deep-synthesis-provisions-2022",
      "cac-ai-content-labeling-measures-2025",
      "cac-national-ai-standardization-guide-2024",
      "cac-algorithm-recommendation-provisions-2022"
    ],
    "confidence_level": "high",
    "last_verified": "2026-07-07",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "For atlas purposes, whether to model the Central Cyberspace Affairs Commission Office and CAC as one node or linked but distinct nodes."
    ],
    "evidence_note_ids": [
      "en-cac-algorithm-recommendation-provisions"
    ]
  },
  {
    "id": "cn-miit",
    "name_en": "Ministry of Industry and Information Technology",
    "name_local": "工业和信息化部",
    "acronym": "MIIT",
    "entity_type": "government",
    "country": "China",
    "city": "Beijing",
    "region": "East Asia",
    "summary": "Central industry and telecom ministry with a major role in the China AI standards and evaluation stack. Primary sources show MIIT co-issuing CAC-facing AI service rules and the 2024 national AI standardization guide, and serving as the home ministry for affiliated institutions including CAICT and CESI.",
    "focus_areas": [
      "ai_governance",
      "standards",
      "policy_regulation",
      "evaluation"
    ],
    "tags": [
      "ministry",
      "standards",
      "co-issuer",
      "industrial policy",
      "industrial AI standardization",
      "telecom and internet administration",
      "co-issuance of AI service rules",
      "ministry-affiliated standards and evaluation institutions",
      "China"
    ],
    "source_ids": [
      "report-china-standards-evaluation-2026-06-05",
      "cac-generative-ai-measures-2023",
      "cac-deep-synthesis-provisions-2022",
      "cac-ai-content-labeling-measures-2025",
      "cac-national-ai-standardization-guide-2024",
      "miit-tc1-homepage"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Whether MIIT/TC1 should be promoted from candidate to full entity in the atlas core graph."
    ]
  },
  {
    "id": "cn-most",
    "name_en": "Ministry of Science and Technology",
    "name_local": "科学技术部",
    "acronym": "MOST",
    "entity_type": "government",
    "country": "China",
    "city": "Beijing",
    "region": "East Asia",
    "summary": "National science and technology ministry. Its clearest sourced AI-governance output in the atlas is the 2019 New Generation AI Governance Principles, an eight-principle responsible-AI framework that includes safe and controllable (安全可控) and agile governance (敏捷治理).",
    "focus_areas": [
      "ai_governance",
      "ethics",
      "policy_regulation"
    ],
    "tags": [
      "ministry",
      "AI governance principles",
      "responsible AI",
      "China"
    ],
    "source_ids": [
      "most-new-generation-ai-governance-principles-2019"
    ],
    "confidence_level": "medium",
    "last_verified": "2026-07-07",
    "tier": 2,
    "needs_primary_source": true,
    "open_questions": [
      "Which MOST units currently hold AI governance responsibilities after the science and technology institutional restructuring.",
      "Whether MOST's 2021 AI ethics norms and the science and technology ethics review measures should be added once primary source pages are verified."
    ],
    "evidence_note_ids": [
      "en-most-2019-governance-principles"
    ]
  },
  {
    "id": "cn-tc260",
    "name_en": "National Cybersecurity Standardization Technical Committee",
    "name_local": "全国网络安全标准化技术委员会",
    "acronym": "TC260",
    "entity_type": "standards_body",
    "country": "China",
    "city": "Beijing",
    "region": "East Asia",
    "website": "https://www.tc260.org.cn/",
    "summary": "China's main national cybersecurity standardization committee with an increasingly visible AI-security and AI-governance role. Official sources show AI-specific standards and guidance on generative AI service security, content labeling methods, AI safety governance, AI application ethics, and the 2026 formation of WG9 for AI security standards.",
    "focus_areas": [
      "standards",
      "model_security",
      "evaluation",
      "cybersecurity",
      "ai_governance",
      "content_safety"
    ],
    "tags": [
      "standards committee",
      "AI security",
      "guidance",
      "WG9",
      "AI security standards",
      "generative AI service security",
      "content labeling methods",
      "AI safety governance framework",
      "AI ethics and application guidance",
      "cybersecurity standardization",
      "China"
    ],
    "source_ids": [
      "report-china-standards-evaluation-2026-06-05",
      "tc260-homepage",
      "tc260-committee-roster-2024",
      "tc260-genai-service-security-requirements-2024",
      "tc260-content-labeling-practice-guide-2023",
      "tc260-ai-safety-governance-framework-2024",
      "shlab-wg9-formation-2026",
      "gb-45438-2025-aigc-labeling-standard"
    ],
    "confidence_level": "high",
    "last_verified": "2026-07-07",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "A dedicated TC260 source explicitly naming the secretariat institution would strengthen administrative-link mapping.",
      "Future updates should distinguish draft, technical-file, and national-standard status more granularly."
    ],
    "evidence_note_ids": [
      "en-gb-45438-labeling-standard"
    ]
  },
  {
    "id": "tc28-sc42",
    "name_en": "AI Subcommittee of the National Information Technology Standardization Technical Committee",
    "name_local": "全国信息技术标准化技术委员会人工智能分技术委员会",
    "acronym": "TC28/SC42",
    "entity_type": "standards_committee",
    "country": "China",
    "region": "East Asia",
    "summary": "National information-technology AI subcommittee under the broader TC28 standards structure. Official sources show CESI as the secretariat unit and confirm the committee's role in AI standardization work.",
    "focus_areas": [
      "standards",
      "evaluation",
      "testing_infrastructure"
    ],
    "tags": [
      "standards committee",
      "AI subcommittee",
      "national standards",
      "AI standards development",
      "national information-technology standardization",
      "committee coordination",
      "AI terminology and methods",
      "China",
      "CESI secretariat"
    ],
    "source_ids": [
      "report-china-standards-evaluation-2026-06-05",
      "tc28-sc42-samr-org-page",
      "tc28-sc42-cesi-page"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "A fuller source on current working groups or current standard projects would improve subcommittee mapping."
    ]
  },
  {
    "id": "cn-caict",
    "name_en": "China Academy of Information and Communications Technology",
    "name_local": "中国信息通信研究院",
    "acronym": "CAICT",
    "entity_type": "evaluation_institution",
    "country": "China",
    "city": "Beijing",
    "region": "East Asia",
    "summary": "MIIT-affiliated research institution that is central to China's applied AI governance, benchmarking, and industry-self-regulation stack. Official sources show CAICT's direct MIIT affiliation, its secretary-general-unit role in jointly founding AIIA, its role in launching the AI Risk Management System and AI Safety Benchmark through AIIA, and its Fangsheng benchmarking system with a dedicated safety testing dimension.",
    "focus_areas": [
      "evaluation",
      "testing_infrastructure",
      "standards",
      "ai_assurance",
      "ai_governance",
      "model_security"
    ],
    "tags": [
      "research institute",
      "benchmarking",
      "AIIA",
      "MIIT-affiliated",
      "AI governance research",
      "AI safety benchmarks",
      "large-model benchmarking",
      "industry commitments",
      "AIIA platform operations",
      "policy and standards support",
      "China"
    ],
    "source_ids": [
      "report-china-standards-evaluation-2026-06-05",
      "caict-about-us",
      "caict-aiia-founding-news-2018",
      "caict-ai-risk-governance-report-2024",
      "caict-ai-safety-governance-report-2025",
      "caict-fangsheng-benchmark-report-2024"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "A dedicated current overview page for CAICT's AI institute or safety team structure would sharpen internal-unit mapping."
    ]
  },
  {
    "id": "aiia",
    "name_en": "China AI Industry Development Alliance",
    "name_local": "中国人工智能产业发展联盟",
    "acronym": "AIIA",
    "entity_type": "industry_alliance",
    "country": "China",
    "region": "East Asia",
    "summary": "Needed because multiple official CAICT sources treat AIIA as the umbrella platform through which CAICT launched safety-governance structures, AI safety commitments, and AI safety benchmarks.",
    "focus_areas": [
      "ai_governance",
      "ai_assurance",
      "evaluation",
      "capacity_building"
    ],
    "tags": [
      "Needed because multiple official CAICT sources treat AIIA as the umbrella platform through which CAICT launched safety-governance structures, AI safety commitments, and AI safety benchmarks.",
      "AIIA",
      "industry alliance",
      "AI safety benchmark"
    ],
    "source_ids": [
      "report-china-standards-evaluation-2026-06-05",
      "caict-aiia-founding-news-2018",
      "caict-ai-risk-governance-report-2024",
      "caict-ai-safety-governance-report-2025"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false
  },
  {
    "id": "aiia-safety-governance-committee",
    "name_en": "AIIA Safety Governance Committee",
    "name_local": "中国人工智能产业发展联盟安全治理委员会",
    "acronym": "AIIA Safety Governance Committee",
    "entity_type": "working_group",
    "country": "China",
    "region": "East Asia",
    "summary": "Best-evidenced through official CAICT reports rather than a dedicated committee page. Official CAICT sources state that CAICT, relying on AIIA, prepared a Safety Governance Committee in late 2023, released an AI Risk Management System, and later linked AIIA to AI safety commitments and the AI Safety Benchmark.",
    "focus_areas": [
      "ai_governance",
      "ai_assurance",
      "evaluation",
      "capacity_building"
    ],
    "tags": [
      "industry committee",
      "AIIA-linked",
      "risk management",
      "committee-in-formation-history",
      "industry safety governance",
      "AI risk management",
      "AI safety benchmark coordination",
      "industry commitments",
      "China",
      "AIIA"
    ],
    "source_ids": [
      "report-china-standards-evaluation-2026-06-05",
      "caict-ai-risk-governance-report-2024",
      "caict-ai-safety-governance-report-2025"
    ],
    "confidence_level": "medium",
    "last_verified": "2026-06-05",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "Dedicated official committee page and current leadership roster.",
      "Whether the committee has a stable standing structure separate from CAICT's report language of 筹建.",
      "Current membership and working-group breakdown."
    ]
  },
  {
    "id": "cesi",
    "name_en": "China Electronics Standardization Institute",
    "name_local": "中国电子技术标准化研究院",
    "acronym": "CESI",
    "entity_type": "evaluation_institution",
    "country": "China",
    "city": "Beijing",
    "region": "East Asia",
    "summary": "MIIT-affiliated standards institute with a central role in China's AI standards and model-evaluation infrastructure. Official sources place CESI as secretariat host for TC28/SC42 and show CESI releasing both a general large-model evaluation standard and the Qiúsuǒ national-standard benchmark system.",
    "focus_areas": [
      "standards",
      "evaluation",
      "testing_infrastructure",
      "ai_assurance"
    ],
    "tags": [
      "standards institute",
      "MIIT-affiliated",
      "evaluation standards",
      "secretariat",
      "AI standardization",
      "committee secretariat functions",
      "large-model evaluation standards",
      "benchmark systems",
      "electronics and information standards",
      "China"
    ],
    "source_ids": [
      "report-china-standards-evaluation-2026-06-05",
      "cesi-basic-info",
      "tc28-sc42-samr-org-page",
      "tc28-sc42-cesi-page",
      "cesi-general-large-model-evaluation-standard-2024",
      "cesi-qiusuo-benchmark-2024"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Whether CESI's benchmark assets and testing services should be modeled as separate nodes or as institute-level capabilities."
    ]
  },
  {
    "id": "cstc",
    "name_en": "China Software Testing Center",
    "name_local": "中国软件评测中心",
    "acronym": "CSTC",
    "entity_type": "evaluation_institution",
    "country": "China",
    "region": "East Asia",
    "summary": "Official site evidence supports CSTC as an AI assessment actor rather than a pure standards drafter. The homepage lists AI evaluation and AIM AI management maturity assessment services, and the center's Sadie Lab activity stream includes an AI large-model safety-governance seminar and AI safety workgroup launch.",
    "focus_areas": [
      "evaluation",
      "testing_infrastructure",
      "ai_assurance",
      "model_security"
    ],
    "tags": [
      "testing center",
      "assessment",
      "AIM",
      "evaluation services",
      "AI evaluation services",
      "AI management maturity assessment",
      "digital-government assessment",
      "AI safety eventing and workgroup activity",
      "China"
    ],
    "source_ids": [
      "report-china-standards-evaluation-2026-06-05",
      "cstc-homepage-ai-evaluation",
      "cstc-ai-big-model-safety-workgroup-2024"
    ],
    "confidence_level": "medium",
    "last_verified": "2026-06-05",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "A directly retrievable official center-introduction page would improve mapping of CSTC's administrative status.",
      "The exact scope and permanence of the cited AI safety workgroup remain unclear from the retrieved snippet."
    ]
  },
  {
    "id": "cn-shlab",
    "name_en": "Shanghai Artificial Intelligence Laboratory",
    "name_local": "上海人工智能实验室",
    "acronym": "SHLAB",
    "entity_type": "research_group",
    "country": "China",
    "city": "Shanghai",
    "region": "East Asia",
    "summary": "Major Shanghai AI research institution with visible roles in evaluation infrastructure, AI security standardization, and PR2G-imported frontier-risk framework work with Concordia AI.",
    "focus_areas": [
      "technical_safety_research",
      "evaluation",
      "testing_infrastructure",
      "model_security",
      "ai_assurance",
      "frontier_ai_safety",
      "dangerous_capability_evaluation"
    ],
    "tags": [
      "research lab",
      "OpenCompass",
      "WG9",
      "evaluation",
      "open model evaluation",
      "AI security standards",
      "medical model testing",
      "trustworthy AI",
      "research infrastructure",
      "China",
      "Concordia framework co-issuer",
      "frontier risk framework"
    ],
    "source_ids": [
      "report-technical-research-groups-2026-06-05",
      "shlab-about-us",
      "shlab-opencompass-2023",
      "shlab-wg9-formation-2026",
      "shanghai-medical-llm-testing-center-2025",
      "concordia-framework"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "Further sources on how SHLAB's governance research center is institutionally nested within SHLAB would help entity-graph precision."
    ],
    "evidence_note_ids": [
      "ev-16-concordia-framework"
    ]
  },
  {
    "id": "shanghai-ai-safety-and-governance-laboratory",
    "name_en": "Shanghai AI Safety and Governance Laboratory",
    "name_local": "上海人工智能安全治理实验室",
    "acronym": "Shanghai AI Safety and Governance Laboratory",
    "entity_type": "research_group",
    "country": "China",
    "city": "Shanghai",
    "region": "East Asia",
    "summary": "Officially evidenced through a government-domain repost from Cyberspace Shanghai. The retrieved source states the lab was jointly established by SHLAB's Governance Research Center and the Shanghai Information Security Testing and Certification Center and unveiled at WAIC 2024.",
    "focus_areas": [
      "ai_governance",
      "technical_safety_research",
      "model_security",
      "ai_assurance"
    ],
    "tags": [
      "local laboratory",
      "Shanghai",
      "governance",
      "jointly established",
      "AI safety governance",
      "Shanghai local experimentation",
      "governance research",
      "safety support ecosystem",
      "China"
    ],
    "source_ids": [
      "report-china-standards-evaluation-2026-06-05",
      "shanghai-ai-safety-governance-lab-unveiling-2024",
      "shlab-about-us"
    ],
    "confidence_level": "medium",
    "last_verified": "2026-06-12",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "A directly retrievable first-party page from the lab itself or from Cyberspace Shanghai.",
      "Current operating team, permanent website, and concrete programs."
    ]
  },
  {
    "id": "tc260-wg9",
    "name_en": "AI Security Standards Working Group",
    "name_local": "人工智能安全标准工作组",
    "acronym": "WG9",
    "entity_type": "working_group",
    "country": "China",
    "region": "East Asia",
    "summary": "Official SHLAB reporting attributes WG9's formal establishment to a TC260 notice and identifies SHLAB's director as group head, making WG9 a material node in the China AI security standards stack.",
    "focus_areas": [
      "standards",
      "model_security",
      "cybersecurity",
      "evaluation"
    ],
    "tags": [
      "Official SHLAB reporting attributes WG9's formal establishment to a TC260 notice and identifies SHLAB's director as group head, making WG9 a material node in the China AI security standards stack.",
      "WG9",
      "AI security standards"
    ],
    "source_ids": [
      "report-china-standards-evaluation-2026-06-05",
      "shlab-wg9-formation-2026",
      "tc260-homepage"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false
  },
  {
    "id": "opencompass",
    "name_en": "OpenCompass at Shanghai AI Laboratory",
    "name_local": "OpenCompass 司南",
    "acronym": "OpenCompass",
    "entity_type": "evaluation_platform",
    "country": "China",
    "region": "East Asia",
    "summary": "Open evaluation infrastructure from Shanghai AI Lab that explicitly includes safety as a model capability dimension.",
    "focus_areas": [
      "evaluation",
      "testing_infrastructure",
      "ai_assurance",
      "open_model_governance"
    ],
    "tags": [
      "Official SHLAB sources present OpenCompass as a SHLAB-built open evaluation system spanning multiple domains, including safety and trustworthiness.",
      "OpenCompass",
      "evaluation platform",
      "open benchmarking",
      "multimodal evaluation",
      "Shanghai AI Lab",
      "model_evaluation",
      "safety_evaluation",
      "technical_testing_infrastructure"
    ],
    "source_ids": [
      "report-technical-research-groups-2026-06-05",
      "shlab-opencompass-2023",
      "src-opencompass-launch",
      "src-opencompass-doc"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Need more public documentation on OpenCompass safety-only tracks and datasets."
    ]
  },
  {
    "id": "cn-baai",
    "name_en": "Beijing Academy of Artificial Intelligence",
    "name_local": "北京智源人工智能研究院",
    "acronym": "BAAI",
    "entity_type": "research_group",
    "country": "China",
    "city": "Beijing",
    "region": "East Asia",
    "summary": "Non-profit new-type AI R&D institution that is significant in China's model-evaluation and AI-safety dialogue ecosystem. Official sources show BAAI maintaining FlagEval as an open evaluation platform and hosting the Beijing AI Safety International Dialogue that produced the Beijing AI Safety Consensus.",
    "focus_areas": [
      "evaluation",
      "technical_safety_research",
      "open_model_governance",
      "international_dialogue"
    ],
    "tags": [
      "research institute",
      "FlagEval",
      "AI safety dialogue",
      "non-profit",
      "large-model evaluation",
      "open research infrastructure",
      "Chinese model ecosystem benchmarking",
      "China"
    ],
    "source_ids": [
      "report-technical-research-groups-2026-06-05",
      "baai-about-us",
      "baai-ai-safety-ecosystem-page",
      "baai-flageval-system-page"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "Whether to model the Beijing AI Safety International Dialogue and Beijing AI Safety Consensus as separate event and document entities."
    ]
  },
  {
    "id": "flag-eval",
    "name_en": "FlagEval",
    "name_local": "FlagEval 天秤大模型评测体系及开放平台",
    "acronym": "FlagEval",
    "entity_type": "evaluation_platform",
    "country": "China",
    "region": "East Asia",
    "summary": "BAAI’s public evaluation platform for large and foundation models, including safety and values dimensions.",
    "focus_areas": [
      "evaluation",
      "testing_infrastructure",
      "open_model_governance",
      "ai_assurance"
    ],
    "tags": [
      "Official BAAI system materials identify FlagEval as BAAI's open model-evaluation platform, making it a useful evaluation-stack node.",
      "FlagEval",
      "large-model evaluation",
      "benchmarking",
      "leaderboard",
      "foundation models",
      "model_evaluation",
      "safety_evaluation",
      "technical_testing_infrastructure"
    ],
    "source_ids": [
      "report-technical-research-groups-2026-06-05",
      "baai-flageval-system-page",
      "src-flageval"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Need current public detail on its safety-specific testing modules."
    ]
  },
  {
    "id": "cn-beijing-aisi",
    "name_en": "Beijing Institute of AI Safety and Governance",
    "name_local": "北京前瞻人工智能安全与治理研究院",
    "acronym": "Beijing AISI",
    "entity_type": "ai_safety_institute",
    "country": "China",
    "city": "Beijing",
    "region": "East Asia",
    "summary": "Beijing's dedicated AI safety-and-governance institute appears to have evolved from a 2024 laboratory phase into a formally established institute in April 2025. Official sources also show a department structure that explicitly distinguishes AI Safety from AI Security and includes guardrails, ethics, and risk-response work.",
    "focus_areas": [
      "frontier_ai_safety",
      "evaluation",
      "ai_governance",
      "red_teaming",
      "model_security"
    ],
    "tags": [
      "local institute",
      "Beijing",
      "AI safety",
      "AI security",
      "AI safety research",
      "AI security research",
      "guardrails",
      "risk response",
      "AI ethics",
      "governance research",
      "AISI",
      "China"
    ],
    "source_ids": [
      "report-china-standards-evaluation-2026-06-05",
      "beijing-ai-safety-lab-launch-2024",
      "beijing-ai-safety-institute-official-launch-2025",
      "beijing-ai-safety-institute-departments"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "How the institute relates administratively to Beijing municipal authorities and other Beijing AI institutions.",
      "Whether the earlier lab name should remain as an alias or a historical predecessor entity."
    ]
  },
  {
    "id": "cn-cnaisda",
    "name_en": "China AI Development and Safety Network",
    "name_local": "中国人工智能发展与安全研究网络",
    "acronym": "CnAISDA",
    "entity_type": "dialogue_convenor",
    "country": "China",
    "region": "East Asia",
    "summary": "Official Chinese institutional coverage positions CnAISDA primarily as a China-side dialogue network corresponding to an AI safety institute for international engagement, not automatically as a domestic frontier-model testing regulator. Tsinghua and SQZ sources explicitly identify CAICT, BAAI, and SHLAB as member units.",
    "focus_areas": [
      "international_dialogue",
      "dialogue_channel",
      "capacity_building",
      "ai_governance"
    ],
    "tags": [
      "network",
      "international dialogue",
      "China-side representative body",
      "not-yet-demonstrated domestic evaluator",
      "international AI safety dialogue",
      "research-network coordination",
      "global governance engagement",
      "cross-institution representation",
      "China",
      "Track 2"
    ],
    "source_ids": [
      "report-china-west-dialogue-2026-06-05",
      "cnaisda-tsinghua-paris-side-event-2025",
      "cnaisda-shanghai-qizhi-side-event-2025"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 3,
    "needs_primary_source": false,
    "open_questions": [
      "Permanent governance structure and secretariat.",
      "Whether future official sources will assign domestic testing or audit functions."
    ]
  },
  {
    "id": "miit-tc1",
    "name_en": "AI Standardization Technical Committee of the Ministry of Industry and Information Technology",
    "name_local": "工业和信息化部人工智能标准化技术委员会",
    "acronym": "MIIT/TC1",
    "entity_type": "standards_committee",
    "country": "China",
    "region": "East Asia",
    "summary": "The official committee homepage shows a distinct ministry-level AI standardization architecture with dedicated working groups and a CAICT-linked operational base, making it highly relevant to the China standards stack even though it was not in the starting cluster.",
    "focus_areas": [
      "standards",
      "policy_regulation",
      "ai_governance"
    ],
    "tags": [
      "The official committee homepage shows a distinct ministry-level AI standardization architecture with dedicated working groups and a CAICT-linked operational base, making it highly relevant to the China standards stack even though it was not in the starting cluster.",
      "MIIT/TC1",
      "standards committee"
    ],
    "source_ids": [
      "report-china-standards-evaluation-2026-06-05",
      "miit-tc1-homepage",
      "miit-ai-standardization-committee-announcement-2024"
    ],
    "confidence_level": "medium",
    "last_verified": "2026-06-12",
    "tier": 1,
    "needs_primary_source": false
  },
  {
    "id": "jp-aisi",
    "name_en": "AI Safety Institute Japan",
    "name_local": "AIセーフティ・インスティテュート",
    "acronym": "AISI Japan",
    "entity_type": "ai_safety_institute",
    "country": "Japan",
    "region": "East Asia",
    "summary": "Strong public technical-evaluation node with explicit guides for LLM-system safety evaluation and red teaming.",
    "focus_areas": [
      "frontier_ai_safety",
      "evaluation",
      "testing_infrastructure",
      "ai_assurance",
      "international_dialogue",
      "dangerous_capability_evaluation",
      "red_teaming",
      "model_security"
    ],
    "tags": [
      "AISI",
      "Japan",
      "large-model evaluation",
      "japan",
      "aisi",
      "evaluation",
      "testing",
      "evaluation methods",
      "guidance",
      "testing tools",
      "evaluation environment",
      "international coordination",
      "evaluation guide",
      "red teaming",
      "LLM systems",
      "safety_evaluation",
      "dangerous-capability_evaluation",
      "jailbreak_or_prompt-injection_evaluation",
      "technical_testing_infrastructure"
    ],
    "source_ids": [
      "report-japan-korea-singapore-aisi-2026-06-05",
      "jpn-aisi-about-page",
      "jpn-integrated-innovation-strategy-2024",
      "jpn-aisi-activity-paper-2025-03",
      "jpn-aisi-eval-tool-oss-post",
      "jpn-aisi-eval-env-taskforce-report-post",
      "src-japan-aisi-eval",
      "src-japan-aisi-redteam"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Whether Japan AISI will move into formal assurance or conformity assessment.",
      "Need public benchmark and dataset outputs beyond guidance documents."
    ]
  },
  {
    "id": "kr-aisi",
    "name_en": "AI Safety Institute Korea",
    "name_local": "인공지능안전연구소",
    "acronym": "AISI Korea",
    "entity_type": "ai_safety_institute",
    "country": "South Korea",
    "region": "East Asia",
    "summary": "Major Korean public evaluation node with explicit risk categories, benchmark infrastructure, and public model-evaluation activity.",
    "focus_areas": [
      "frontier_ai_safety",
      "evaluation",
      "testing_infrastructure",
      "policy_regulation",
      "technical_safety_research",
      "international_dialogue",
      "dangerous_capability_evaluation",
      "red_teaming",
      "model_security"
    ],
    "tags": [
      "AISI",
      "Korea",
      "large-model evaluation",
      "south-korea",
      "aisi",
      "policy",
      "evaluation",
      "research",
      "risk definition",
      "evaluation frameworks",
      "evaluation infrastructure",
      "international cooperation",
      "risk categories",
      "42-model program",
      "safety_evaluation",
      "dangerous-capability_evaluation",
      "deception",
      "jailbreak_or_prompt-injection_evaluation",
      "technical_testing_infrastructure"
    ],
    "source_ids": [
      "report-japan-korea-singapore-aisi-2026-06-05",
      "kor-aisi-vision-mission",
      "kor-aisi-evaluation-page",
      "kor-msit-launch-release",
      "kor-etri-aisi-page",
      "kor-aisi-consortium-meeting-post",
      "src-korea-aisi-home",
      "src-korea-aisi-42models"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "How Korea AISI will divide future work between evaluation, policy support, and standards-adjacent functions.",
      "Need public names and links for specific datasets and tools hosted by the institute."
    ]
  },
  {
    "id": "sg-aisi",
    "name_en": "Singapore AI Safety Institute",
    "name_local": "Singapore AI Safety Institute",
    "acronym": "AISI",
    "entity_type": "ai_safety_institute",
    "country": "Singapore",
    "region": "Southeast Asia",
    "summary": "Publicly structured as a designation-plus-partnership model rather than a standalone statutory institute: DTC is the designated AISI and IMDA is the overall government agency.",
    "focus_areas": [
      "frontier_ai_safety",
      "evaluation",
      "testing_infrastructure",
      "ai_assurance",
      "international_dialogue",
      "policy_regulation"
    ],
    "tags": [
      "AISI",
      "Singapore",
      "large-model evaluation",
      "singapore",
      "aisi",
      "testing",
      "evaluation",
      "policy",
      "tools",
      "research",
      "international testing leadership",
      "designation-plus-partnership model"
    ],
    "source_ids": [
      "report-japan-korea-singapore-aisi-2026-06-05",
      "sg-aisi-about-page",
      "sg-aisi-work-page",
      "sg-aisi-resources-page"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Whether Singapore AISI will become a more standalone institution."
    ]
  },
  {
    "id": "sg-ai-verify-foundation",
    "name_en": "AI Verify Foundation",
    "name_local": "AI Verify Foundation",
    "acronym": "AIVF",
    "entity_type": "evaluation_institution",
    "country": "Singapore",
    "region": "Southeast Asia",
    "summary": "Singapore’s strongest public technical-testing ecosystem node, relevant here because it operates Project Moonshot for LLM benchmarking and red teaming.",
    "focus_areas": [
      "ai_assurance",
      "evaluation",
      "testing_infrastructure",
      "standards",
      "open_model_governance"
    ],
    "tags": [
      "AI Verify",
      "Singapore",
      "model card",
      "singapore",
      "aivf",
      "assurance",
      "tooling",
      "frameworks",
      "sandboxing",
      "accreditation",
      "Project Moonshot",
      "sandbox",
      "benchmarking",
      "red teaming",
      "AI assurance",
      "testing ecosystem",
      "open-source",
      "safety_evaluation",
      "technical_testing_infrastructure",
      "open-source_or_open-weight_model_governance"
    ],
    "source_ids": [
      "report-japan-korea-singapore-aisi-2026-06-05",
      "report-asean-genai-governance-2026-06-05",
      "sg-ai-verify-foundation-about",
      "sg-ai-verify-testing-framework-page",
      "sg-project-moonshot-page",
      "sg-global-ai-assurance-sandbox-page",
      "sg-ai-tester-accreditation-page",
      "ai-verify-foundation-home-2026",
      "global-ai-assurance-sandbox-main-report-2025",
      "src-aivf-moonshot-launch"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "How AIVF and Singapore AISI will divide future public-facing evaluation roles.",
      "Publicly documented governance and scope details for the AI Tester Accreditation Programme remain limited in the sources captured here.",
      "Need a public breakdown of Moonshot modules and released case studies."
    ]
  },
  {
    "id": "asean-working-group-ai-governance",
    "name_en": "ASEAN Working Group on AI Governance",
    "acronym": "WG-AI",
    "entity_type": "working_group",
    "country": "Regional",
    "region": "Southeast Asia",
    "summary": "ASEAN regional working group on AI governance. The Bangkok Digital Declaration directly supports a coordination remit that includes AI policy, governance, safety, laws, regulations, and standards, but does not establish supranational regulatory authority.",
    "focus_areas": [
      "ai_governance",
      "policy_regulation",
      "capacity_building",
      "international_dialogue"
    ],
    "tags": [
      "ASEAN",
      "AI governance",
      "capacity building",
      "WG-AI",
      "regional coordination"
    ],
    "source_ids": [
      "report-asean-genai-governance-2026-06-05",
      "bangkok-digital-declaration-2025",
      "expanded-asean-guide-genai-2025"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-12",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Official terms of reference and current membership for WG-AI."
    ]
  },
  {
    "id": "org-concordia-ai",
    "name_en": "Concordia AI",
    "entity_type": "think_tank",
    "country": "International",
    "region": "Cross-border",
    "summary": "Social enterprise with offices in Beijing and Singapore. PR2G treats Concordia as a highly APAC-linked comparator across China/Singapore research, standards and policy engagement, and frontier-risk framework work with Shanghai AI Laboratory.",
    "focus_areas": [
      "international_dialogue",
      "ai_governance",
      "technical_safety_research",
      "frontier_ai_safety",
      "standards",
      "evaluation",
      "dangerous_capability_evaluation",
      "biosecurity",
      "cybersecurity",
      "open_source_governance"
    ],
    "tags": [
      "Track 2",
      "China-West dialogue",
      "technical research",
      "Concordia AI",
      "Beijing office",
      "Singapore office",
      "China report",
      "Singapore report",
      "frontier risk framework",
      "standards participation"
    ],
    "source_ids": [
      "report-china-west-dialogue-2026-06-05",
      "concordia-home",
      "concordia-china-report",
      "concordia-singapore-report",
      "concordia-policy",
      "concordia-framework"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "Specific WAIC event claims remain held until an accessible official event source is available."
    ],
    "evidence_note_ids": [
      "ev-15-concordia-home",
      "ev-15-concordia-china-report",
      "ev-15-concordia-singapore-report",
      "ev-15-concordia-policy",
      "ev-16-concordia-framework"
    ]
  },
  {
    "id": "org-safe-ai-forum",
    "name_en": "Safe AI Forum",
    "entity_type": "dialogue_convenor",
    "country": "International",
    "region": "Cross-border",
    "summary": "Non-profit dialogue convenor whose flagship series is IDAIS. PR2G imports it as directly APAC-relevant through Beijing and London IDAIS dialogue architecture.",
    "focus_areas": [
      "international_dialogue",
      "dialogue_channel",
      "frontier_ai_safety",
      "ai_governance",
      "capacity_building"
    ],
    "tags": [
      "Track 2",
      "dialogue channel",
      "frontier AI safety",
      "Safe AI Forum",
      "IDAIS",
      "dialogue convenor",
      "China-linked dialogue"
    ],
    "source_ids": [
      "report-china-west-dialogue-2026-06-05",
      "saif",
      "farai-idais-page"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "Session-specific source coverage is incomplete for IDAIS-Oxford and IDAIS-Venice in this payload."
    ],
    "evidence_note_ids": [
      "ev-13-saif",
      "ev-13-idais-home"
    ]
  },
  {
    "id": "event-idais",
    "name_en": "International Dialogues on AI Safety",
    "acronym": "IDAIS",
    "entity_type": "event",
    "country": "International",
    "region": "Cross-border",
    "summary": "Recurring international AI safety dialogue series. PR2G strengthens the record with normalized SAIF/IDAIS evidence and imports Beijing/London session events where source IDs are available.",
    "focus_areas": [
      "international_dialogue",
      "dialogue_channel",
      "frontier_ai_safety"
    ],
    "tags": [
      "Track 2",
      "dialogue channel",
      "IDAIS",
      "Safe AI Forum flagship series"
    ],
    "source_ids": [
      "report-china-west-dialogue-2026-06-05",
      "farai-idais-page",
      "idais-beijing"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 3,
    "needs_primary_source": false,
    "open_questions": [
      "Which session-specific IDAIS pages should anchor China-side participant claims?",
      "Oxford and Venice session source rows remain held until explicit source records are available in the import payload."
    ],
    "evidence_note_ids": [
      "ev-13-idais-home",
      "ev-14-idais-beijing",
      "ev-14-idais-home"
    ]
  },
  {
    "id": "cn-tsinghua-iaiig",
    "name_en": "Tsinghua Institute for AI International Governance",
    "name_local": "清华大学人工智能国际治理研究院",
    "acronym": "Tsinghua I-AIIG",
    "entity_type": "university",
    "country": "China",
    "city": "Beijing",
    "region": "East Asia",
    "summary": "Tsinghua-linked AI international governance research node included for policy, governance, and cross-border dialogue mapping.",
    "focus_areas": [
      "ai_governance",
      "international_dialogue",
      "policy_regulation"
    ],
    "tags": [
      "Track 2",
      "China",
      "AI governance"
    ],
    "source_ids": [
      "report-technical-research-groups-2026-06-05",
      "report-china-west-dialogue-2026-06-05"
    ],
    "confidence_level": "low",
    "last_verified": "2026-06-12",
    "tier": 2,
    "needs_primary_source": true
  },
  {
    "id": "cn-tsinghua-ciss",
    "name_en": "Tsinghua Center for International Security and Strategy",
    "name_local": "清华大学战略与安全研究中心",
    "acronym": "CISS",
    "entity_type": "think_tank",
    "country": "China",
    "city": "Beijing",
    "region": "East Asia",
    "summary": "Tsinghua-linked center that appears in public records for U.S.-China and China-EU AI and international security dialogues. PR2C keeps these descriptions high-level and models the rows as public dialogue channels, not operational security relationships.",
    "focus_areas": [
      "international_dialogue",
      "dialogue_channel",
      "ai_governance",
      "policy_regulation"
    ],
    "tags": [
      "Track 2",
      "China",
      "international security",
      "CISS",
      "track 2 dialogue",
      "China-EU AI and international security dialogue",
      "U.S.-China AI and international security dialogue"
    ],
    "source_ids": [
      "report-china-west-dialogue-2026-06-05",
      "ciss-china-eu-ai-security-dialogue-round-2-2024",
      "ciss-china-eu-ai-security-dialogue-round-5-2026",
      "brookings-ai-glossary-2024"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Additional exact public source paths would strengthen the seventh U.S.-China dialogue row before import."
    ]
  },
  {
    "id": "ipa-japan",
    "name_en": "Information-technology Promotion Agency, Japan",
    "name_local": "独立行政法人 情報処理推進機構",
    "acronym": "IPA",
    "entity_type": "government",
    "country": "Japan",
    "region": "East Asia",
    "summary": "Host agency in which Japan AISI is placed and where the AISI secretariat was set up.",
    "focus_areas": [
      "ai_governance",
      "capacity_building",
      "testing_infrastructure"
    ],
    "tags": [
      "japan",
      "ipa",
      "host-agency",
      "hosting",
      "administrative support"
    ],
    "source_ids": [
      "jpn-aisi-about-page",
      "jpn-integrated-innovation-strategy-2024"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Public sources do not fully specify AISI’s operational autonomy inside IPA."
    ]
  },
  {
    "id": "cabinet-office-aisi-liaison-japan",
    "name_en": "AISI Related Ministries and Agencies Liaison Meeting",
    "name_local": "AIセーフティ・インスティテュート関係府省庁等連絡会議",
    "entity_type": "policy_program",
    "country": "Japan",
    "region": "East Asia",
    "summary": "Cabinet Office-linked inter-ministerial coordination mechanism around Japan AISI, with ministries, research institutes, IPA, and AISI represented.",
    "focus_areas": [
      "ai_governance",
      "policy_regulation",
      "international_dialogue",
      "capacity_building"
    ],
    "tags": [
      "japan",
      "cabinet-office",
      "coordination",
      "inter-ministerial linkage",
      "inter-ministerial coordination",
      "liaison meeting"
    ],
    "source_ids": [
      "jpn-cabinet-office-aisi-liaison-setup",
      "jpn-aisi-about-page"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-12",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Its precise decision authority is not clear in reviewed public sources."
    ]
  },
  {
    "id": "msit-korea",
    "name_en": "Ministry of Science and ICT",
    "name_local": "과학기술정보통신부",
    "acronym": "MSIT",
    "entity_type": "government",
    "country": "South Korea",
    "region": "East Asia",
    "summary": "The ministry that launched Korea AISI and publicly ties it to post-Seoul-Summit implementation and international AI safety networking.",
    "focus_areas": [
      "ai_governance",
      "policy_regulation",
      "international_dialogue"
    ],
    "tags": [
      "south-korea",
      "msit",
      "ministry",
      "institution creation",
      "AI safety policy",
      "international engagement"
    ],
    "source_ids": [
      "kor-msit-launch-release",
      "kor-msit-tta-nist-crosswalk-release"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "How MSIT oversight and resourcing of Korea AISI will evolve over time."
    ]
  },
  {
    "id": "etri-korea",
    "name_en": "Electronics and Telecommunications Research Institute",
    "name_local": "한국전자통신연구원",
    "acronym": "ETRI",
    "entity_type": "research_group",
    "country": "South Korea",
    "region": "East Asia",
    "summary": "The clearest public institutional home for Korea AISI in reviewed sources.",
    "focus_areas": [
      "technical_safety_research",
      "evaluation",
      "testing_infrastructure"
    ],
    "tags": [
      "south-korea",
      "etri",
      "research-institute",
      "hosting",
      "research support",
      "institutional home"
    ],
    "source_ids": [
      "kor-etri-aisi-page",
      "kor-aisi-vision-mission"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-12",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Whether Korea AISI will remain embedded in ETRI long term."
    ]
  },
  {
    "id": "korea-ai-safety-consortium",
    "name_en": "Korea AI Safety Consortium",
    "name_local": "AI안전컨소시엄",
    "entity_type": "industry_alliance",
    "country": "South Korea",
    "region": "East Asia",
    "summary": "Consortium formed at Korea AISI’s launch to support framework work, subgroup collaboration, and joint-testing discussions.",
    "focus_areas": [
      "capacity_building",
      "evaluation",
      "international_dialogue",
      "ai_governance"
    ],
    "tags": [
      "south-korea",
      "consortium",
      "coordination",
      "subgroups",
      "joint testing discussion"
    ],
    "source_ids": [
      "kor-msit-launch-release",
      "kor-aisi-consortium-meeting-post"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "The full signed member list was not confirmed in reviewed primary sources."
    ]
  },
  {
    "id": "digital-trust-centre-singapore",
    "name_en": "Digital Trust Centre Singapore",
    "name_local": "Digital Trust Centre",
    "acronym": "DTC",
    "entity_type": "research_group",
    "country": "Singapore",
    "region": "Southeast Asia",
    "summary": "NTU-based centre designated as Singapore AISI and serving as the technical backbone of the Singapore model.",
    "focus_areas": [
      "technical_safety_research",
      "ai_assurance",
      "testing_infrastructure",
      "capacity_building"
    ],
    "tags": [
      "singapore",
      "dtc",
      "ntu",
      "research-centre",
      "technical work",
      "research ecosystem coordination",
      "talent",
      "trust-tech R&D",
      "NTU",
      "technical backbone"
    ],
    "source_ids": [
      "sg-aisi-about-page",
      "sg-dtc-home-page"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "How DTC’s wider digital-trust mandate will balance with narrower AISI-style functions."
    ]
  },
  {
    "id": "imda-singapore",
    "name_en": "Infocomm Media Development Authority",
    "name_local": "Infocomm Media Development Authority",
    "acronym": "IMDA",
    "entity_type": "government",
    "country": "Singapore",
    "region": "Southeast Asia",
    "summary": "Update IMDA with Singapore’s assurance and testing implementation roles, including A.I. Verify launch, the Global AI Assurance Sandbox with AI Verify Foundation, the Singapore AI Safety Red Teaming Challenge, and 2026 testing-standardisation activity.",
    "focus_areas": [
      "ai_governance",
      "policy_regulation",
      "ai_assurance",
      "international_dialogue",
      "evaluation",
      "testing_infrastructure"
    ],
    "tags": [
      "singapore",
      "imda",
      "government-agency",
      "assurance",
      "policy",
      "international engagement",
      "assurance ecosystem building",
      "baseline policy",
      "A.I. Verify",
      "red teaming",
      "sandbox",
      "standards"
    ],
    "source_ids": [
      "sg-aisi-about-page",
      "sg-imda-llm-starter-kit",
      "imda-ai-verify-launch-2022",
      "sg-global-ai-assurance-sandbox-page",
      "singapore-ai-safety-red-teaming-challenge-page-2025",
      "singapore-global-ai-testing-standardisation-2026"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Public sources reviewed do not fully separate IMDA’s AISI role from its broader trusted-AI policy role.",
      "Whether IMDA has published a final formal specification for an AI tester accreditation programme beyond current public pages.",
      "Whether Global AI Assurance Sandbox participation criteria or reports are updated on a fixed cadence."
    ]
  },
  {
    "id": "ai-verify-testing-framework",
    "name_en": "AI Verify Testing Framework",
    "name_local": "AI Verify Testing Framework",
    "entity_type": "policy_program",
    "country": "Singapore",
    "region": "Asia-Pacific",
    "summary": "Core assurance framework in Singapore’s public stack; should be separated from Singapore AISI itself.",
    "focus_areas": [
      "ai_assurance",
      "evaluation",
      "standards"
    ],
    "tags": [
      "AI Verify",
      "testing framework",
      "AI assurance",
      "framework"
    ],
    "source_ids": [
      "sg-ai-verify-testing-framework-page",
      "sg-ai-verify-foundation-about"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-12",
    "tier": 1,
    "needs_primary_source": false
  },
  {
    "id": "project-moonshot",
    "name_en": "Project Moonshot",
    "name_local": "Project Moonshot",
    "acronym": "Moonshot",
    "entity_type": "evaluation_platform",
    "country": "Singapore",
    "region": "Southeast Asia",
    "summary": "Open-source LLM evaluation toolkit that combines benchmarking and red teaming for LLMs and LLM applications.",
    "focus_areas": [
      "evaluation",
      "testing_infrastructure",
      "red_teaming",
      "ai_assurance",
      "model_security"
    ],
    "tags": [
      "Project Moonshot",
      "open-source toolkit",
      "LLM evaluation",
      "red-teaming",
      "tool",
      "LLM toolkit",
      "benchmarking",
      "red teaming",
      "safety_evaluation",
      "jailbreak_or_prompt-injection_evaluation",
      "technical_testing_infrastructure"
    ],
    "source_ids": [
      "sg-project-moonshot-page",
      "sg-imda-llm-starter-kit",
      "expanded-asean-guide-genai-2025",
      "src-aivf-moonshot-launch"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Need public inventory of datasets, threat classes, and supported report outputs."
    ]
  },
  {
    "id": "global-ai-assurance-sandbox",
    "name_en": "Global AI Assurance Sandbox",
    "name_local": "Global AI Assurance Sandbox",
    "entity_type": "policy_program",
    "country": "Singapore",
    "region": "Asia-Pacific",
    "summary": "Singapore-linked assurance programme for deployed or deploying GenAI applications, represented as application-level assurance and market-building infrastructure rather than foundation-model safety testing.",
    "focus_areas": [
      "ai_assurance",
      "capacity_building",
      "evaluation",
      "testing_infrastructure"
    ],
    "tags": [
      "AI assurance",
      "sandbox",
      "IMDA",
      "AIVF",
      "programme",
      "application testing"
    ],
    "source_ids": [
      "sg-global-ai-assurance-sandbox-page",
      "global-ai-assurance-sandbox-main-report-2025"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false
  },
  {
    "id": "ai-tester-accreditation-programme",
    "name_en": "AI Tester Accreditation Programme",
    "name_local": "AI Tester Accreditation Programme",
    "acronym": "AI TAP",
    "entity_type": "policy_program",
    "country": "Singapore",
    "region": "Asia-Pacific",
    "summary": "Distinct third-party testing accreditation programme in Singapore’s assurance ecosystem.",
    "focus_areas": [
      "ai_assurance",
      "capacity_building",
      "evaluation",
      "testing_infrastructure"
    ],
    "tags": [
      "AI TAP",
      "tester accreditation",
      "third-party testing",
      "programme"
    ],
    "source_ids": [
      "sg-ai-tester-accreditation-page",
      "ai-verify-foundation-home-2026"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-12",
    "tier": 1,
    "needs_primary_source": false
  },
  {
    "id": "centre-for-humanitarian-dialogue",
    "name_en": "Centre for Humanitarian Dialogue",
    "acronym": "HD",
    "entity_type": "dialogue_convenor",
    "country": "Switzerland",
    "region": "Europe",
    "summary": "Public co-host institution for repaired China-EU AI and International Security Dialogue round records. The atlas models this as a dialogue-convening role only.",
    "focus_areas": [
      "international_dialogue",
      "dialogue_channel",
      "ai_governance"
    ],
    "tags": [
      "HD Centre",
      "China-EU dialogue",
      "track 2 dialogue",
      "dialogue convenor"
    ],
    "source_ids": [
      "ciss-china-eu-ai-security-dialogue-round-2-2024",
      "ciss-china-eu-ai-security-dialogue-round-5-2026"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "A more specific HD institutional page would improve the organization profile; PR2C only imports the event co-host evidence."
    ]
  },
  {
    "id": "shanghai-qi-zhi-institute",
    "name_en": "Shanghai Qi Zhi Institute",
    "name_local": "上海期智研究院",
    "entity_type": "research_group",
    "country": "China",
    "city": "Shanghai",
    "region": "East Asia",
    "summary": "China-side research institute visible in this PR2C import through the source-backed CnAISDA Paris side-event record. Broader IDAIS-related claims are held until exact URLs are repaired.",
    "focus_areas": [
      "international_dialogue",
      "ai_governance"
    ],
    "tags": [
      "SQZ",
      "CnAISDA",
      "Paris side event",
      "China-facing interface"
    ],
    "source_ids": [
      "cnaisda-shanghai-qizhi-side-event-2025",
      "shanghai-qizhi-institute-introduction-2026"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-12",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "Exact source URLs are still needed before importing the IDAIS Oxfordshire relationship rows."
    ]
  },
  {
    "id": "deepseek",
    "name_en": "DeepSeek",
    "name_local": "深度求索",
    "acronym": "DeepSeek",
    "entity_type": "company",
    "country": "China",
    "region": "East Asia",
    "summary": "Chinese frontier-model developer with public disclosures centered on a transparency page plus open-weight DeepSeek-V3 repository and license documentation. This backfill does not identify a public preparedness policy, dangerous-capability evaluation framework, or external frontier-safety audit.",
    "focus_areas": [
      "frontier_model_developer",
      "open_model_governance",
      "model_security",
      "content_safety"
    ],
    "tags": [
      "disclosure:transparency_channel",
      "disclosure:model_card",
      "disclosure:custom_license",
      "gap:no_public_rsp",
      "gap:no_public_dangerous_capability_eval",
      "gap:no_public_external_audit",
      "frontier_model_development",
      "open_weight_release",
      "api_platform",
      "source_type_mapping:repository_to_technical_report",
      "source_type_mapping:license_to_policy_document",
      "platform disclosure",
      "model license",
      "PR2D: public disclosure backfill"
    ],
    "source_ids": [
      "deepseek-transparency-center",
      "deepseek-v3-github",
      "deepseek-v3-license"
    ],
    "confidence_level": "medium",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Is there any public DeepSeek API safety or abuse-policy document beyond the transparency hub?",
      "Are there public dangerous-capability or release-gating evaluations for frontier DeepSeek models?",
      "Is there any public third-party red-team or external audit report?"
    ]
  },
  {
    "id": "alibaba-qwen",
    "name_en": "Alibaba Qwen",
    "name_local": "阿里巴巴通义千问",
    "acronym": "Qwen",
    "entity_type": "company",
    "country": "China",
    "region": "East Asia",
    "summary": "Alibaba Qwen model-family record backed in this pass by official repository and model-card documentation. The public evidence supports open-weight/model-card disclosures, not a public dangerous-capability or release-gating framework.",
    "focus_areas": [
      "frontier_model_developer",
      "open_model_governance",
      "evaluation"
    ],
    "tags": [
      "disclosure:model_card",
      "disclosure:repository_release",
      "disclosure:open_weight_release",
      "gap:no_public_rsp",
      "gap:no_public_dangerous_capability_eval",
      "gap:platform_policy_docs_incomplete",
      "frontier_model_development",
      "open_weight_release",
      "api_platform",
      "source_type_mapping:repository_to_technical_report",
      "source_type_mapping:model_card_to_technical_report",
      "model card",
      "repository release",
      "PR2D: public disclosure backfill"
    ],
    "source_ids": [
      "qwen3-github-readme",
      "qwen3-model-card"
    ],
    "confidence_level": "medium",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "What official Alibaba/Qwen privacy, service-agreement, or platform-guardrail pages should be added for a fuller ba...",
      "Are there official Qwen red-teaming, dangerous-capability, or release-gating disclosures?",
      "Is there any explicit public advanced-AI risk statement tied to Qwen releases?"
    ]
  },
  {
    "id": "zhipu-ai-glm",
    "name_en": "Zhipu AI",
    "name_local": "智谱AI",
    "acronym": "GLM",
    "entity_type": "company",
    "country": "China",
    "region": "East Asia",
    "summary": "Chinese frontier-model developer whose BigModel documentation surfaces API access, content-safety controls, model-evaluation tooling, GLM-4.5 model information, and privacy policy material. These are platform and evaluation disclosures, not a public preparedness framework.",
    "focus_areas": [
      "frontier_model_developer",
      "evaluation",
      "content_safety",
      "model_security"
    ],
    "tags": [
      "disclosure:api_controls",
      "disclosure:content_safety",
      "disclosure:model_evaluation",
      "disclosure:alignment_training",
      "disclosure:agentic_model_docs",
      "gap:no_public_rsp",
      "gap:no_public_dangerous_capability_eval",
      "frontier_model_development",
      "api_platform",
      "model_evaluation",
      "BigModel",
      "platform controls",
      "privacy",
      "model evaluation",
      "content safety",
      "PR2D: public disclosure backfill"
    ],
    "source_ids": [
      "zhipu-api-overview",
      "zhipu-content-safety",
      "zhipu-model-evaluation",
      "zhipu-glm-4-5-overview",
      "zhipu-privacy-policy"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Does Zhipu publish model-specific red-team results for frontier GLM releases?",
      "Is there any public external audit or third-party safety evaluation?",
      "Are there any public deployment thresholds or release-gating criteria beyond the general platform-evaluation tooling?"
    ]
  },
  {
    "id": "baidu-ernie",
    "name_en": "Baidu ERNIE",
    "name_local": "百度文心",
    "acronym": "ERNIE",
    "entity_type": "company",
    "country": "China",
    "region": "East Asia",
    "summary": "Baidu ERNIE/Qianfan record backed by Qianfan platform, user-agreement, intervention, API, and content-marking documentation. Public disclosures are strongest on platform governance, content controls, data handling, and generated-content labeling.",
    "focus_areas": [
      "frontier_model_developer",
      "evaluation",
      "content_safety",
      "model_security",
      "ai_governance"
    ],
    "tags": [
      "disclosure:user_data_handling",
      "disclosure:content_safety",
      "disclosure:api_controls",
      "disclosure:content_provenance",
      "disclosure:model_evaluation_platform",
      "gap:no_public_rsp",
      "gap:no_public_dangerous_capability_eval",
      "frontier_model_development",
      "api_platform",
      "enterprise_ai_platform",
      "Qianfan",
      "privacy",
      "content marking",
      "intervention rules",
      "API controls",
      "PR2D: public disclosure backfill"
    ],
    "source_ids": [
      "baidu-qianfan-overview",
      "baidu-qianfan-user-agreement",
      "baidu-qianfan-intervention-strategy",
      "baidu-qianfan-api-general",
      "baidu-qianfan-content-marking"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Is there a public ERNIE model card or technical report for the latest frontier closed models with safety sections?",
      "Are any dangerous-capability evaluations or external red-team exercises publicly disclosed?",
      "Are there model-specific release gates distinct from Qianfan platform controls?"
    ]
  },
  {
    "id": "tencent-hunyuan",
    "name_en": "Tencent Hunyuan",
    "name_local": "腾讯混元",
    "acronym": "Hunyuan",
    "entity_type": "company",
    "country": "China",
    "region": "East Asia",
    "summary": "Tencent Hunyuan record backed by Tencent Cloud product, API, compatibility, moderation, and image-labeling documentation. The evidence supports platform controls and alignment-language disclosures, not a public frontier-risk governance policy.",
    "focus_areas": [
      "frontier_model_developer",
      "content_safety",
      "model_security",
      "ai_governance"
    ],
    "tags": [
      "disclosure:api_controls",
      "disclosure:content_safety",
      "disclosure:content_provenance",
      "disclosure:alignment_language",
      "disclosure:model_lineup",
      "gap:no_public_rsp",
      "gap:no_public_dangerous_capability_eval",
      "frontier_model_development",
      "api_platform",
      "multimodal_models",
      "Hunyuan",
      "API controls",
      "image moderation",
      "content provenance",
      "human preference alignment language",
      "PR2D: public disclosure backfill"
    ],
    "source_ids": [
      "tencent-hunyuan-product-overview",
      "tencent-hunyuan-product-dynamics",
      "tencent-hunyuan-api-key-management",
      "tencent-hunyuan-openai-compat",
      "tencent-hunyuan-image-faq"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Are there public Hunyuan model cards or technical reports for the latest closed frontier releases with safety sections?",
      "Is any external red-teaming or third-party audit public?",
      "Do public materials disclose release gates for advanced reasoning/agent models?"
    ]
  },
  {
    "id": "bytedance-doubao",
    "name_en": "ByteDance Doubao",
    "name_local": "字节跳动豆包",
    "acronym": "Doubao",
    "entity_type": "company",
    "country": "China",
    "region": "East Asia",
    "summary": "ByteDance/Doubao record backed by Volcengine Ark and Doubao service documentation. Public materials emphasize platform terms, API access, and model-evaluation workflow disclosures; snippet-only platform-scope claims were held out of evidence and relationships.",
    "focus_areas": [
      "frontier_model_developer",
      "evaluation",
      "model_security",
      "ai_governance"
    ],
    "tags": [
      "disclosure:platform_security",
      "disclosure:api_controls",
      "disclosure:model_evaluation_platform",
      "disclosure:service_terms",
      "gap:no_public_rsp",
      "gap:no_public_dangerous_capability_eval",
      "frontier_model_development",
      "api_platform",
      "agent_platform",
      "Doubao",
      "Volcengine Ark",
      "API platform",
      "model evaluation tooling",
      "service terms",
      "PR2D: public disclosure backfill"
    ],
    "source_ids": [
      "volcengine-ark-overview",
      "volcengine-ark-special-terms",
      "volcengine-doubao-model-service-agreement",
      "volcengine-api-key-config",
      "volcengine-model-evaluation"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Are there public Doubao model cards or technical reports for the newest frontier closed models?",
      "Are there public dangerous-capability or misuse evaluations beyond generic model evaluation tooling?",
      "Is there any public external audit or red-team disclosure for Doubao/Ark?"
    ]
  },
  {
    "id": "moonshot-kimi",
    "name_en": "Moonshot AI Kimi",
    "name_local": "月之暗面 Kimi",
    "acronym": "Kimi",
    "entity_type": "company",
    "country": "China",
    "region": "East Asia",
    "summary": "Moonshot AI Kimi record backed by Kimi privacy, service-agreement, platform, and Kimi K2 model-card materials. Public evidence supports privacy, developer data-responsibility, tooling, and model-card disclosures rather than release gates or advanced-risk governance.",
    "focus_areas": [
      "frontier_model_developer",
      "model_security",
      "ai_governance"
    ],
    "tags": [
      "disclosure:privacy",
      "disclosure:developer_data_responsibility",
      "disclosure:tooling",
      "disclosure:model_card",
      "gap:no_public_rsp",
      "gap:no_public_dangerous_capability_eval",
      "gap:no_public_external_audit",
      "frontier_model_development",
      "api_platform",
      "agentic_models",
      "Kimi",
      "privacy",
      "developer data responsibility",
      "model card",
      "tooling",
      "PR2D: public disclosure backfill"
    ],
    "source_ids": [
      "kimi-api-platform-home",
      "kimi-privacy-policy",
      "kimi-service-agreement",
      "kimi-k2-instruct-model-card"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Does Moonshot publish any model-specific safety card or red-team summary for current frontier releases?",
      "Are there public release gates or preparedness thresholds for agentic Kimi models?",
      "Is any external audit or third-party model-safety evaluation public?"
    ]
  },
  {
    "id": "minimax",
    "name_en": "MiniMax",
    "name_local": "MiniMax",
    "acronym": "MiniMax",
    "entity_type": "company",
    "country": "China",
    "region": "East Asia",
    "summary": "MiniMax record backed by about, API, privacy, model-page, and content-labeling materials. Public disclosures are mostly API/platform, privacy, benchmark, and content-provenance oriented; snippet-only API-key and labeling evidence notes were held.",
    "focus_areas": [
      "frontier_model_developer",
      "content_safety",
      "model_security",
      "open_model_governance",
      "evaluation"
    ],
    "tags": [
      "disclosure:api_controls",
      "disclosure:privacy",
      "disclosure:content_provenance",
      "disclosure:model_benchmarks",
      "gap:no_public_rsp",
      "gap:no_public_dangerous_capability_eval",
      "gap:no_public_external_audit",
      "frontier_model_development",
      "api_platform",
      "multimodal_models",
      "open_weight_release",
      "MiniMax API",
      "privacy",
      "content labeling",
      "model benchmarks",
      "multimodal models",
      "PR2D: public disclosure backfill"
    ],
    "source_ids": [
      "minimax-about",
      "minimax-api-prerequisites",
      "minimax-api-key-faq",
      "minimax-api-privacy-policy",
      "minimax-m3-model-page",
      "minimax-content-labeling-announcement"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Is there any public MiniMax model-card-style safety section for frontier text models?",
      "Does MiniMax disclose red-teaming or external evaluations for advanced agentic/coding releases?",
      "Are there public release-gating or post-deployment monitoring criteria beyond API and labeling controls?"
    ]
  },
  {
    "id": "stepfun",
    "name_en": "StepFun",
    "name_local": "阶跃星辰",
    "acronym": "Step",
    "entity_type": "company",
    "country": "China",
    "region": "East Asia",
    "summary": "StepFun record backed by privacy, user-agreement, platform, migration, and reasoning API pages. Because PR2D evidence for detailed API and agent-framework claims was snippet-only, this import keeps StepFun claims at the general platform-disclosure level.",
    "focus_areas": [
      "frontier_model_developer",
      "model_security",
      "ai_governance"
    ],
    "tags": [
      "disclosure:privacy",
      "disclosure:api_controls",
      "disclosure:agent_framework_compatibility",
      "disclosure:model_routing",
      "gap:no_public_rsp",
      "gap:no_public_dangerous_capability_eval",
      "frontier_model_development",
      "api_platform",
      "agentic_models",
      "StepFun platform",
      "privacy",
      "API compatibility",
      "reasoning models",
      "snippet-only detailed claims held",
      "PR2D: public disclosure backfill"
    ],
    "source_ids": [
      "stepfun-privacy-policy",
      "stepfun-open-platform-home",
      "stepfun-user-agreement",
      "stepfun-openai-migration",
      "step-plan-reasoning-api"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Are there public StepFun model cards or technical reports for Step 2/3.x frontier models with safety sections?",
      "Are any external red-team or dangerous-capability evaluations public?",
      "Is there a public release-gating or preparedness document distinct from API/platform docs?"
    ]
  },
  {
    "id": "zero-one-ai-yi",
    "name_en": "01.AI",
    "name_local": "零一万物",
    "acronym": "Yi",
    "entity_type": "company",
    "country": "China",
    "region": "East Asia",
    "summary": "01.AI/Yi record backed by repository, model-card, and licensing source rows. This pass treats Yi as primarily an open-weight documentation case and does not import snippet-only claims about training details, context windows, or license language.",
    "focus_areas": [
      "frontier_model_developer",
      "open_model_governance"
    ],
    "tags": [
      "disclosure:model_card",
      "disclosure:open_weight_release",
      "disclosure:apache2_license",
      "gap:no_public_platform_privacy_docs_found",
      "gap:no_public_rsp",
      "gap:no_public_dangerous_capability_eval",
      "frontier_model_development",
      "open_weight_release",
      "coding_models",
      "Yi",
      "open-weight release",
      "model card",
      "license",
      "snippet-only detail claims held",
      "PR2D: public disclosure backfill"
    ],
    "source_ids": [
      "yi-1-5-github",
      "yi-1-5-34b-chat-model-card",
      "yi-license"
    ],
    "confidence_level": "medium",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "What is 01.AI's current official API/privacy/service-agreement stack for Yi if any public platform remains acti...",
      "Are there any official red-team or dangerous-capability evaluation disclosures for Yi?",
      "Is there any public advanced-AI-risk or preparedness statement from 01.AI?"
    ]
  },
  {
    "id": "zhipu-bigmodel-open-platform",
    "name_en": "BigModel Open Platform",
    "name_local": "大模型开放平台",
    "acronym": "BigModel",
    "entity_type": "policy_program",
    "country": "China",
    "region": "Asia-Pacific",
    "summary": "Official Zhipu platform surface used in PR2D for source-backed API access, content-safety, and model-evaluation disclosures.",
    "focus_areas": [
      "evaluation",
      "content_safety",
      "model_security"
    ],
    "tags": [
      "platform",
      "company-operated platform",
      "PR2D platform node"
    ],
    "source_ids": [
      "zhipu-api-overview",
      "zhipu-content-safety",
      "zhipu-model-evaluation"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false
  },
  {
    "id": "baidu-qianfan-platform",
    "name_en": "Baidu Qianfan Platform",
    "name_local": "百度千帆·大模型服务及Agent开发平台",
    "acronym": "Qianfan",
    "entity_type": "policy_program",
    "country": "China",
    "region": "Asia-Pacific",
    "summary": "Official Baidu Qianfan platform surface used in PR2D for model services, intervention rules, API access, and content-marking controls.",
    "focus_areas": [
      "evaluation",
      "content_safety",
      "model_security",
      "ai_governance"
    ],
    "tags": [
      "platform",
      "company-operated platform",
      "PR2D platform node"
    ],
    "source_ids": [
      "baidu-qianfan-overview",
      "baidu-qianfan-intervention-strategy",
      "baidu-qianfan-content-marking"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false
  },
  {
    "id": "kimi-api-platform",
    "name_en": "Kimi API Open Platform",
    "name_local": "Kimi API 开放平台",
    "acronym": "Kimi API",
    "entity_type": "policy_program",
    "country": "China",
    "region": "Asia-Pacific",
    "summary": "Official Moonshot AI Kimi API platform surface used in PR2D for source-backed tooling, privacy, service-agreement, and model-card disclosures.",
    "focus_areas": [
      "model_security",
      "ai_governance"
    ],
    "tags": [
      "platform",
      "company-operated platform",
      "PR2D platform node"
    ],
    "source_ids": [
      "kimi-api-platform-home",
      "kimi-privacy-policy",
      "kimi-service-agreement"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-05",
    "tier": 1,
    "needs_primary_source": false
  },
  {
    "id": "asean-secretariat",
    "name_en": "Association of Southeast Asian Nations",
    "acronym": "ASEAN",
    "entity_type": "dialogue_convenor",
    "country": "Regional",
    "region": "Southeast Asia",
    "summary": "Regional intergovernmental organization represented here only for source-backed ASEAN AI governance coordination: the 2024 voluntary AI governance guide, the 2025 generative-AI expansion, the Bangkok Digital Declaration, and a Malaysia official ASEAN AI Safe announcement. Held roadmap and search-snippet claims are not treated as operational evidence.",
    "focus_areas": [
      "ai_governance",
      "policy_regulation",
      "international_dialogue",
      "capacity_building"
    ],
    "tags": [
      "ASEAN",
      "regional voluntary guidance",
      "political declaration",
      "AI Safe",
      "ADGMIN"
    ],
    "source_ids": [
      "asean-guide-ai-governance-ethics-2024",
      "expanded-asean-guide-genai-2025",
      "bangkok-digital-declaration-2025",
      "malaysia-champions-asean-ai-safe-2025"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Whether ASEAN AI Safe has published terms of reference, secretariat arrangements, or a public work programme beyond summit-level announcements.",
      "Whether the Responsible AI Roadmap has a final public PDF working text beyond the official page summary captured here."
    ]
  },
  {
    "id": "mydigital-corporation",
    "name_en": "MyDIGITAL Corporation",
    "acronym": "MyDIGITAL",
    "entity_type": "government",
    "country": "Malaysia",
    "region": "Southeast Asia",
    "summary": "Malaysian government-linked implementation and coordination node in this source pack, connected to NAIO incubation, AI action-plan procurement, and a Malaysia official announcement about ASEAN AI Safe. The public sources do not establish MyDIGITAL as an AI safety testing body.",
    "focus_areas": [
      "ai_governance",
      "policy_regulation",
      "international_dialogue",
      "capacity_building"
    ],
    "tags": [
      "NAIO incubation",
      "AI action plan",
      "ASEAN AI Safe",
      "Malaysia"
    ],
    "source_ids": [
      "mydigital-naio-page",
      "mydigital-national-ai-action-plan-rfp-2025",
      "malaysia-champions-asean-ai-safe-2025"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Whether NAIO’s long-term institutional home remains within MyDIGITAL or shifted after incubation."
    ]
  },
  {
    "id": "etda-thailand",
    "name_en": "Electronic Transactions Development Agency",
    "name_local": "สำนักงานพัฒนาธุรกรรมทางอิเล็กทรอนิกส์",
    "acronym": "ETDA",
    "entity_type": "government",
    "country": "Thailand",
    "region": "Southeast Asia",
    "summary": "Thai digital-governance agency represented here through official AI governance guidance, AI Governance Center activity, ethics principles, and draft AI-law consultation sources. The imported rows do not treat the draft-law consultation as enacted law.",
    "focus_areas": [
      "ai_governance",
      "ethics",
      "policy_regulation",
      "capacity_building"
    ],
    "tags": [
      "AIGC",
      "AI law draft",
      "AI governance guideline",
      "Thailand"
    ],
    "source_ids": [
      "etda-aigc-page",
      "thailand-ai-governance-guideline-executive-2023",
      "digital-thailand-ai-ethics-guideline-2019",
      "etda-draft-ai-law-pr-2025"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Whether Thailand has published the full draft AI law text or only public hearing principles in an official accessible form."
    ]
  },
  {
    "id": "dost-asti-ph",
    "name_en": "Advanced Science and Technology Institute",
    "acronym": "DOST-ASTI",
    "entity_type": "research_group",
    "country": "Philippines",
    "region": "Southeast Asia",
    "summary": "Philippine government research institute represented here as a public AI R&D implementation actor, including DOST-ASTI AI programme materials, SkAI-Pinas-linked work, and the launch of NAICRI.",
    "focus_areas": [
      "technical_safety_research",
      "testing_infrastructure",
      "capacity_building",
      "ai_governance"
    ],
    "tags": [
      "NAICRI",
      "SkAI-Pinas",
      "AI R&D",
      "Philippines"
    ],
    "source_ids": [
      "philippines-dost-asti-ai-page",
      "philippines-naicri-launch-2026",
      "philippines-dost-ai-flagship-year1-2026"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Whether DOST or DICT has separately published a final Philippines National AI Strategy Roadmap 2.0 text."
    ]
  },
  {
    "id": "brin-orei",
    "name_en": "Electronics and Informatics Research Organization",
    "name_local": "Organisasi Riset Elektronika dan Informatika",
    "acronym": "OREI",
    "entity_type": "research_group",
    "country": "Indonesia",
    "region": "Southeast Asia",
    "summary": "BRIN research organization represented with limited public evidence that it hosts an AI and cybersecurity research center page. Details on mandate, staffing, and programme scope remain sparse in this source pack.",
    "focus_areas": [
      "technical_safety_research",
      "cybersecurity"
    ],
    "tags": [
      "BRIN",
      "research center",
      "AI and cybersecurity",
      "Indonesia"
    ],
    "source_ids": [
      "brin-ai-cybersecurity-research-center-page-2025",
      "brin-orei-home-2026"
    ],
    "confidence_level": "medium",
    "last_verified": "2026-06-12",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "Formal mandate, staffing, and programme portfolio of the BRIN AI and Cybersecurity Research Center.",
      "Whether BRIN or another Indonesian ministry hosts the authoritative public Stranas KA implementation portal."
    ]
  },
  {
    "id": "asean-ai-safe-network",
    "name_en": "ASEAN AI Safety Network",
    "acronym": "ASEAN AI Safe",
    "entity_type": "policy_program",
    "country": "Regional",
    "region": "Southeast Asia",
    "summary": "Regional AI safety network initiative supported here by a Malaysia official announcement. The atlas does not treat it as an operational testing network until ASEAN publishes public implementation details.",
    "focus_areas": [
      "ai_governance",
      "international_dialogue",
      "capacity_building"
    ],
    "tags": [
      "ASEAN AI Safe",
      "regional network",
      "political declaration"
    ],
    "source_ids": [
      "malaysia-champions-asean-ai-safe-2025",
      "expanded-asean-guide-genai-2025"
    ],
    "confidence_level": "medium",
    "last_verified": "2026-06-12",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Public terms of reference, membership rules, secretariat arrangements, or work programme for ASEAN AI Safe."
    ]
  },
  {
    "id": "singapore-ai-safety-red-teaming-challenge",
    "name_en": "Singapore AI Safety Red Teaming Challenge",
    "entity_type": "evaluation_platform",
    "country": "Singapore",
    "region": "Asia-Pacific",
    "summary": "IMDA-organized multicultural and multilingual GenAI red-teaming challenge focused on harmful bias stereotypes across Asian contexts.",
    "focus_areas": [
      "red_teaming",
      "evaluation",
      "ai_assurance",
      "testing_infrastructure",
      "content_safety"
    ],
    "tags": [
      "red teaming",
      "challenge",
      "multilingual testing",
      "Asian contexts"
    ],
    "source_ids": [
      "singapore-ai-safety-red-teaming-challenge-page-2025",
      "singapore-ai-safety-red-teaming-challenge-report-2025"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-12",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Whether IMDA will publish repeat challenge methodology or ongoing programme governance."
    ]
  },
  {
    "id": "national-ai-office-malaysia",
    "name_en": "National AI Office",
    "acronym": "NAIO",
    "entity_type": "government",
    "country": "Malaysia",
    "region": "Southeast Asia",
    "summary": "Malaysia national AI coordination node launched in December 2024. Sources describe policy and regulatory framework shaping, innovation, adoption, and talent functions, but do not establish statutory enforcement powers.",
    "focus_areas": [
      "ai_governance",
      "policy_regulation",
      "capacity_building"
    ],
    "tags": [
      "NAIO",
      "central authority",
      "coordination",
      "Malaysia"
    ],
    "source_ids": [
      "mydigital-naio-page",
      "naio-home-working-groups-page"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Whether NAIO’s formal authority is later codified in statute, regulation, or cabinet decision."
    ]
  },
  {
    "id": "naio-working-groups-malaysia",
    "name_en": "NAIO Working Groups",
    "acronym": "NAIO WGs",
    "entity_type": "working_group",
    "country": "Malaysia",
    "region": "Southeast Asia",
    "summary": "Set of NAIO specialised working groups publicly announced to drive Malaysia’s AI agenda with technology, academic, industry, government, and civil-society participation.",
    "focus_areas": [
      "ai_governance",
      "policy_regulation",
      "capacity_building"
    ],
    "tags": [
      "working groups",
      "multi-stakeholder",
      "Malaysia"
    ],
    "source_ids": [
      "naio-home-working-groups-page",
      "mydigital-naio-page"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-12",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "Current roster and published outputs for each NAIO working group."
    ]
  },
  {
    "id": "ai-governance-center-thailand",
    "name_en": "AI Governance Center",
    "name_local": "ศูนย์ธรรมาภิบาลปัญญาประดิษฐ์",
    "acronym": "AIGC",
    "entity_type": "government",
    "country": "Thailand",
    "region": "Southeast Asia",
    "summary": "ETDA-hosted AI governance center framed as a cross-network cooperation platform for governance framework development, consultation, knowledge exchange, and networking.",
    "focus_areas": [
      "ai_governance",
      "policy_regulation",
      "capacity_building",
      "international_dialogue"
    ],
    "tags": [
      "AIGC",
      "ETDA",
      "Thailand",
      "governance center"
    ],
    "source_ids": [
      "etda-aigc-page",
      "thailand-ai-governance-guideline-executive-2023"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-12",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Whether AIGC has a public service catalogue or formal assessment mandate beyond governance consultation and knowledge exchange."
    ]
  },
  {
    "id": "ai-thailand-program",
    "name_en": "AI Thailand",
    "entity_type": "policy_program",
    "country": "Thailand",
    "region": "Southeast Asia",
    "summary": "Official national AI strategy and action-plan programme page for Thailand, including readiness work on ethics, law and regulation and a strategy line for AI sandboxes.",
    "focus_areas": [
      "ai_governance",
      "policy_regulation",
      "capacity_building",
      "testing_infrastructure"
    ],
    "tags": [
      "national strategy",
      "sandbox",
      "Thailand"
    ],
    "source_ids": [
      "ai-thailand-national-strategy-2022-2027-page",
      "digital-thailand-ai-ethics-guideline-2019"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-12",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "Operational status and public outputs of the AI sandbox elements described in the strategy page."
    ]
  },
  {
    "id": "brin-ai-cybersecurity-research-center",
    "name_en": "AI and Cybersecurity Research Center",
    "name_local": "Pusat Riset Kecerdasan Artifisial dan Keamanan Siber",
    "entity_type": "research_group",
    "country": "Indonesia",
    "region": "Southeast Asia",
    "summary": "BRIN page indicating a dedicated AI and cybersecurity research center under the Electronics and Informatics research organization. Public detail in this source pack is limited.",
    "focus_areas": [
      "technical_safety_research",
      "cybersecurity"
    ],
    "tags": [
      "BRIN",
      "AI and cybersecurity",
      "research center"
    ],
    "source_ids": [
      "brin-ai-cybersecurity-research-center-page-2025",
      "brin-orei-home-2026"
    ],
    "confidence_level": "medium",
    "last_verified": "2026-06-12",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "Mandate, staffing, research outputs, and service portfolio."
    ]
  },
  {
    "id": "naicri-philippines",
    "name_en": "National Artificial Intelligence Center for Research and Innovation",
    "acronym": "NAICRI",
    "entity_type": "research_group",
    "country": "Philippines",
    "region": "Southeast Asia",
    "summary": "Philippine national institutional anchor for AI research, advanced computing, and innovation launched by DOST-ASTI in February 2026.",
    "focus_areas": [
      "technical_safety_research",
      "testing_infrastructure",
      "capacity_building",
      "ai_governance"
    ],
    "tags": [
      "NAICRI",
      "advanced computing",
      "AI research",
      "Philippines"
    ],
    "source_ids": [
      "philippines-naicri-launch-2026",
      "philippines-dost-ai-flagship-year1-2026"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Public service offerings, compute-access rules, and data-governance policies."
    ]
  },
  {
    "id": "skai-pinas",
    "name_en": "Philippine Sky Artificial Intelligence Program",
    "acronym": "SkAI-Pinas",
    "entity_type": "research_group",
    "country": "Philippines",
    "region": "Southeast Asia",
    "summary": "Philippine AI R&D programme represented because DOST-ASTI’s official AI page identifies ASTI-ALaM as part of the broader SkAI-Pinas umbrella.",
    "focus_areas": [
      "technical_safety_research",
      "capacity_building"
    ],
    "tags": [
      "SkAI-Pinas",
      "ASTI-ALaM",
      "research programme"
    ],
    "source_ids": [
      "philippines-dost-asti-ai-page",
      "philippines-dost-ai-flagship-year1-2026"
    ],
    "confidence_level": "medium",
    "last_verified": "2026-06-12",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "Programme charter, governance, and current scope."
    ]
  },
  {
    "id": "pku-alignment-team",
    "name_en": "PKU-Alignment Team",
    "name_local": "北大对齐小组",
    "acronym": "PKU-Alignment",
    "entity_type": "research_group",
    "country": "China",
    "region": "East Asia",
    "summary": "Public PKU team focused on LLM alignment, Safe-RLHF, multimodal alignment, and alignment-brittleness research.",
    "focus_areas": [
      "technical_alignment",
      "evaluation",
      "testing_infrastructure"
    ],
    "tags": [
      "RLHF",
      "multimodal alignment",
      "open-source",
      "safety_evaluation",
      "technical_testing_infrastructure"
    ],
    "source_ids": [
      "src-pku-align-anything",
      "src-pku-brittleness",
      "src-pku-alignment-github",
      "src-pku-safe-rlhf"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Need clearer mapping to the broader PKU Center for AI Safety and Governance."
    ]
  },
  {
    "id": "baai-flagsafe",
    "name_en": "FlagSafe",
    "name_local": "FlagSafe 大模型安全平台",
    "acronym": "FlagSafe",
    "entity_type": "evaluation_platform",
    "country": "China",
    "region": "East Asia",
    "summary": "BAAI-led public safety platform spanning red teaming, defense, interpretability, deception, and alignment.",
    "focus_areas": [
      "technical_alignment",
      "evaluation",
      "dangerous_capability_evaluation",
      "agent_safety",
      "model_security",
      "testing_infrastructure"
    ],
    "tags": [
      "red team",
      "blue team",
      "white-box",
      "Chinese safety platform",
      "safety_evaluation",
      "deception",
      "technical_testing_infrastructure"
    ],
    "source_ids": [
      "src-flagsafe-home",
      "src-flagsafe-launch",
      "src-flagsafe-alignment",
      "src-flagsafe-deception"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Whether FlagSafe should be modeled as a platform or as the public face of a wider BAAI safety center."
    ]
  },
  {
    "id": "ruc-xiting-wang-group",
    "name_en": "Xiting Wang Group at Renmin University of China",
    "name_local": "王诣挺课题组",
    "entity_type": "research_group",
    "country": "China",
    "region": "East Asia",
    "summary": "Publicly described as working on large-model interpretability and safety alignment.",
    "focus_areas": [
      "technical_alignment"
    ],
    "tags": [
      "RUC",
      "interpretability",
      "safety alignment"
    ],
    "source_ids": [
      "src-ruc-xiting-bio"
    ],
    "confidence_level": "medium",
    "last_verified": "2026-06-06",
    "tier": 2,
    "needs_primary_source": true,
    "open_questions": [
      "Need a fuller official lab page tying the group to specific benchmarks or tools."
    ]
  },
  {
    "id": "bigai",
    "name_en": "Beijing Institute for General Artificial Intelligence",
    "name_local": "北京通用人工智能研究院",
    "acronym": "BIGAI",
    "entity_type": "research_group",
    "country": "China",
    "region": "East Asia",
    "summary": "BIGAI is included for TongTest and explicit ability-plus-value-alignment framing; the row remains medium-confidence because the work is broader AGI evaluation rather than pure frontier LLM safety.",
    "focus_areas": [
      "technical_alignment",
      "testing_infrastructure",
      "agent_safety"
    ],
    "tags": [
      "TongTest",
      "AGI evaluation",
      "value alignment",
      "technical_testing_infrastructure"
    ],
    "source_ids": [
      "src-bigai-tongtest"
    ],
    "confidence_level": "medium",
    "last_verified": "2026-06-12",
    "tier": 2,
    "needs_primary_source": true,
    "open_questions": [
      "How much of TongTest now directly targets foundation-model safety."
    ]
  },
  {
    "id": "hkust-shuai-wang-group",
    "name_en": "Shuai Wang Group at HKUST CSE",
    "name_local": "香港科技大学王帅课题组",
    "entity_type": "research_group",
    "country": "Hong Kong",
    "region": "East Asia",
    "summary": "Public HKUST materials show direct LLM security relevance and broader trustworthy-LLM framing around security and reliability risks.",
    "focus_areas": [
      "model_security",
      "cybersecurity"
    ],
    "tags": [
      "LLM security",
      "watermarking",
      "trustworthy LLM",
      "ai-cybersecurity"
    ],
    "source_ids": [
      "src-hkust-shuaiwang",
      "src-hkust-trustworthy-llm"
    ],
    "confidence_level": "medium",
    "last_verified": "2026-06-06",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "Need more official evidence of jailbreak or prompt-injection-specific work."
    ]
  },
  {
    "id": "hong-kong-generative-ai-rd-center",
    "name_en": "Hong Kong Generative AI Research and Development Center",
    "name_local": "香港生成式人工智能研发中心",
    "acronym": "HKGAI",
    "entity_type": "research_group",
    "country": "Hong Kong",
    "region": "East Asia",
    "summary": "Publicly visible through HKGAI-V1, which is framed around Hong Kong-specific alignment, safety, and contextual suitability.",
    "focus_areas": [
      "technical_alignment",
      "evaluation",
      "open_model_governance"
    ],
    "tags": [
      "regional alignment",
      "Hong Kong sovereign model",
      "HKGAI-V1",
      "safety_evaluation",
      "open-source_or_open-weight_model_governance"
    ],
    "source_ids": [
      "src-hkust-hkgai-v1",
      "src-hkgai-v1-paper"
    ],
    "confidence_level": "medium",
    "last_verified": "2026-06-06",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "Need more official center pages on benchmarks and released safety tools."
    ]
  },
  {
    "id": "nii-llmc",
    "name_en": "NII Large Language Model Research and Development Center",
    "name_local": "国立情報学研究所 大規模言語モデル研究開発センター",
    "acronym": "LLMC",
    "entity_type": "research_group",
    "country": "Japan",
    "region": "East Asia",
    "summary": "Key Japanese public node for technical LLM safety, with a Safety WG and safety symposiums.",
    "focus_areas": [
      "technical_alignment",
      "evaluation"
    ],
    "tags": [
      "LLM-jp",
      "symposium",
      "working group",
      "safety_evaluation"
    ],
    "source_ids": [
      "src-nii-llmc",
      "src-nii-today",
      "src-llmjp-resources"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Need a fuller map of LLMC technical-safety outputs beyond talks and reports."
    ]
  },
  {
    "id": "llm-jp-safety-wg",
    "name_en": "LLM-jp Safety Working Group",
    "name_local": "LLM-jp 安全性ワーキンググループ",
    "acronym": "LLM-jp Safety WG",
    "entity_type": "working_group",
    "country": "Japan",
    "region": "East Asia",
    "summary": "Public safety working group inside Japan’s open LLM ecosystem, visible through NII and LLM-jp resources.",
    "focus_areas": [
      "technical_alignment",
      "evaluation"
    ],
    "tags": [
      "LLM-jp",
      "Safety WG",
      "Japan",
      "safety_evaluation"
    ],
    "source_ids": [
      "src-nii-llmc",
      "src-llmjp-resources"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Need clearer public authorship and institutional membership data."
    ]
  },
  {
    "id": "sb-intuitions-responsible-ai-team",
    "name_en": "Responsible AI Team at SB Intuitions",
    "name_local": "Responsible AI チーム",
    "entity_type": "research_group",
    "country": "Japan",
    "region": "East Asia",
    "summary": "Corporate technical-safety team with public Japanese benchmarks for safety-boundary testing and guardrails.",
    "focus_areas": [
      "evaluation",
      "red_teaming",
      "model_security"
    ],
    "tags": [
      "guardrails",
      "Japanese benchmarks",
      "corporate R&D",
      "safety_evaluation",
      "jailbreak_or_prompt-injection_evaluation"
    ],
    "source_ids": [
      "src-sb-safety-boundary",
      "src-sb-wildguard-jp"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Need more public detail on multilingual scope beyond Japanese evaluations."
    ]
  },
  {
    "id": "riken-aip-nlu",
    "name_en": "RIKEN AIP Natural Language Understanding Team",
    "name_local": "理研AIP 自然言語理解チーム",
    "entity_type": "research_group",
    "country": "Japan",
    "region": "East Asia",
    "summary": "Relevant mainly through public interpretability and trustworthy-foundation framing plus adjacent Japanese LLM safety dataset work. This row remains medium-confidence and is not labeled as a core frontier AI safety lab.",
    "focus_areas": [
      "technical_alignment"
    ],
    "tags": [
      "trustworthy NLP",
      "interpretability",
      "AnswerCarefully adjacency"
    ],
    "source_ids": [
      "src-riken-answercarefully",
      "src-riken-nlu"
    ],
    "confidence_level": "medium",
    "last_verified": "2026-06-06",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "Need more direct frontier-model safety outputs from the team itself."
    ]
  },
  {
    "id": "soongsil-ai-safety-center",
    "name_en": "Soongsil University AI Safety Research Center",
    "name_local": "숭실대학교 AI안전성연구센터",
    "acronym": "AISC",
    "entity_type": "research_group",
    "country": "South Korea",
    "region": "East Asia",
    "summary": "University-based Korean group with direct public evidence of large-model security and safety benchmarking using prompt-injection and jailbreak attacks.",
    "focus_areas": [
      "evaluation",
      "red_teaming",
      "model_security",
      "cybersecurity"
    ],
    "tags": [
      "57 attacks",
      "20-model benchmark",
      "foundation-model testing",
      "safety_evaluation",
      "jailbreak_or_prompt-injection_evaluation",
      "ai-cybersecurity"
    ],
    "source_ids": [
      "src-soongsil-center",
      "src-soongsil-llm-safety"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Need released methodology, benchmark artifacts, and scoring details."
    ]
  },
  {
    "id": "kaist-aipr-lab",
    "name_en": "KAIST Artificial Intelligence and Probabilistic Reasoning Lab",
    "name_local": "KAIST AIPR Lab",
    "acronym": "AIPR",
    "entity_type": "research_group",
    "country": "South Korea",
    "region": "East Asia",
    "summary": "KAIST lab with direct public work on jailbreak defense for misaligned instructions and transformer interpretability.",
    "focus_areas": [
      "technical_alignment",
      "red_teaming",
      "model_security"
    ],
    "tags": [
      "GC-DPO",
      "Monet",
      "safe diffusion",
      "interpretability",
      "jailbreak_or_prompt-injection_evaluation"
    ],
    "source_ids": [
      "src-kaist-aipr-home",
      "src-kaist-aipr-pubs"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Need public benchmark or tool releases beyond publications."
    ]
  },
  {
    "id": "smu-jun-sun-group",
    "name_en": "Jun Sun Group at Singapore Management University",
    "name_local": "Jun Sun research group",
    "entity_type": "research_group",
    "country": "Singapore",
    "region": "Southeast Asia",
    "summary": "Publicly relevant through agent-safety, backdoor, and LLM-safety work such as AgentSpec and BackdoorLLM.",
    "focus_areas": [
      "agent_safety",
      "model_security",
      "red_teaming",
      "technical_alignment"
    ],
    "tags": [
      "AgentSpec",
      "BackdoorLLM",
      "SMU",
      "jailbreak_or_prompt-injection_evaluation"
    ],
    "source_ids": [
      "src-smu-agentspec",
      "src-smu-backdoorllm",
      "src-smu-llm-seminar"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Need a fuller public map of the group’s current safety portfolio beyond highlighted papers."
    ]
  },
  {
    "id": "nus-jin-song-dong-group",
    "name_en": "Jin-Song Dong Group at National University of Singapore",
    "name_local": "Jin-Song Dong research group",
    "entity_type": "research_group",
    "country": "Singapore",
    "region": "Southeast Asia",
    "summary": "NUS group relevant through formal methods with LLM agents, trusted AI, and the TrustLLM benchmark.",
    "focus_areas": [
      "technical_alignment",
      "evaluation",
      "agent_safety"
    ],
    "tags": [
      "formal methods",
      "LLM agents",
      "TrustLLM",
      "verified synthesis",
      "safety_evaluation"
    ],
    "source_ids": [
      "src-nus-jinsong",
      "src-trustllm"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Need public clarification of which current projects are benchmarks versus deployed verification tools."
    ]
  },
  {
    "id": "hong-kong-university-of-science-and-technology",
    "name_en": "Hong Kong University of Science and Technology",
    "name_local": "香港科技大学",
    "acronym": "HKUST",
    "entity_type": "university",
    "country": "Hong Kong",
    "region": "East Asia",
    "summary": "University parent institution for source-backed HKUST technical safety and LLM-security rows in PR2F, including Shuai Wang group materials and HKGAI-V1 public activity.",
    "focus_areas": [
      "technical_safety_research",
      "model_security",
      "technical_alignment"
    ],
    "tags": [
      "HKUST",
      "Hong Kong",
      "LLM security",
      "HKGAI"
    ],
    "source_ids": [
      "src-hkust-shuaiwang",
      "src-hkust-hkgai-v1"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "Which HKUST centers or labs own current public safety benchmarks and technical deliverables."
    ]
  },
  {
    "id": "us-caisi",
    "name_en": "Center for AI Standards and Innovation",
    "acronym": "CAISI",
    "entity_type": "ai_safety_institute",
    "country": "United States",
    "region": "North America",
    "summary": "NIST-hosted U.S. comparator entity for AI standards, testing, collaborative research, voluntary agreements, and broad risk-evaluation work. PR2G treats CAISI as an institutional comparator and does not import operational benchmark details.",
    "focus_areas": [
      "frontier_ai_safety",
      "evaluation",
      "dangerous_capability_evaluation",
      "model_security",
      "biosecurity",
      "cybersecurity",
      "standards",
      "testing_infrastructure"
    ],
    "tags": [
      "United States",
      "NIST-hosted",
      "AISI comparator",
      "PRC model evaluation",
      "standards and metrology"
    ],
    "source_ids": [
      "caisi-main",
      "caisi-deepseek"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Which current CAISI outputs should be modeled as research outputs without exposing operational benchmark detail."
    ],
    "evidence_note_ids": [
      "ev-01-caisi-main",
      "ev-02-caisi-deepseek"
    ]
  },
  {
    "id": "uk-aisi",
    "name_en": "UK AI Security Institute",
    "acronym": "UK AISI",
    "entity_type": "ai_safety_institute",
    "country": "United Kingdom",
    "region": "Europe",
    "summary": "State-backed advanced-AI safety and security institute used as a Western comparator for evaluation, mitigation testing, and international coordination, with APAC pathways through Australia and India-linked coalition activity.",
    "focus_areas": [
      "frontier_ai_safety",
      "evaluation",
      "dangerous_capability_evaluation",
      "model_security",
      "technical_safety_research",
      "testing_infrastructure",
      "international_dialogue"
    ],
    "tags": [
      "United Kingdom",
      "AI Security Institute",
      "IN-AIMES",
      "Australia MoU",
      "Alignment Project"
    ],
    "source_ids": [
      "uk-aisi-home",
      "uk-aus-mou",
      "uk-alignment-project"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "Which official source reconciles the IN-AIMES name with older International Network of AI Safety Institutes naming."
    ],
    "evidence_note_ids": [
      "ev-04-uk-aisi-home",
      "ev-05-uk-aus-mou",
      "ev-07-uk-alignment-project"
    ]
  },
  {
    "id": "eu-ai-office",
    "name_en": "European AI Office",
    "entity_type": "government",
    "country": "European Union",
    "region": "Europe",
    "summary": "European Commission office implementing the AI Act, especially for general-purpose AI. It is modeled as a regulatory and evaluation comparator rather than a lab-style AISI.",
    "focus_areas": [
      "ai_governance",
      "policy_regulation",
      "evaluation",
      "standards",
      "international_dialogue",
      "ai_assurance"
    ],
    "tags": [
      "European Commission",
      "AI Act",
      "GPAI",
      "evaluation comparator",
      "international governance"
    ],
    "source_ids": [
      "eu-ai-office",
      "eu-ai-office-establishing-decision-2024"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-12",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "How the AI Office will operationalize GPAI evaluation tools, methodologies, and benchmarks in public materials."
    ],
    "evidence_note_ids": [
      "ev-08-eu-ai-office"
    ]
  },
  {
    "id": "oecd-ai",
    "name_en": "OECD Artificial Intelligence Policy Observatory",
    "acronym": "OECD.AI",
    "entity_type": "standards_body",
    "country": "International",
    "region": "Cross-border",
    "summary": "OECD-operated AI policy observatory hosting the AI Principles, AI Incidents and Hazards Monitor, Hiroshima AI Reporting Framework, and OECD-GPAI materials. It is the strongest standards/reporting comparator in PR2G.",
    "focus_areas": [
      "standards",
      "evaluation",
      "international_dialogue",
      "capacity_building",
      "policy_regulation",
      "dialogue_channel"
    ],
    "tags": [
      "OECD.AI",
      "AI Principles",
      "Hiroshima AI Reporting Framework",
      "AI Incidents Monitor",
      "interoperability"
    ],
    "source_ids": [
      "oecd-ai-home",
      "oecd-ai-principles",
      "oecd-aim",
      "oecd-haip"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "How to connect OECD reporting mechanisms to individual APAC country profiles without implying national implementation."
    ],
    "evidence_note_ids": [
      "ev-09-oecd-ai-home",
      "ev-10-oecd-principles",
      "ev-11-oecd-aim",
      "ev-12-oecd-haip"
    ]
  },
  {
    "id": "nist",
    "name_en": "National Institute of Standards and Technology",
    "acronym": "NIST",
    "entity_type": "standards_body",
    "country": "United States",
    "region": "North America",
    "summary": "U.S. federal standards institution imported only as CAISI’s parent/host and the AI Consortium source for broader AI metrology and standards context.",
    "focus_areas": [
      "standards",
      "evaluation",
      "testing_infrastructure",
      "ai_assurance"
    ],
    "tags": [
      "standards body",
      "metrology",
      "AI Consortium",
      "CAISI parent"
    ],
    "source_ids": [
      "caisi-main",
      "nist-ai-consortium"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "Which NIST standards outputs should be imported separately from CAISI."
    ],
    "evidence_note_ids": [
      "ev-01-caisi-main",
      "ev-03-nist-ai-consortium"
    ]
  },
  {
    "id": "in-aimes",
    "name_en": "International Network for Advanced AI Measurement, Evaluation and Science",
    "acronym": "IN-AIMES",
    "entity_type": "working_group",
    "country": "International",
    "region": "Cross-border",
    "summary": "Current official UK government naming found in PR2G for an international AISI-style measurement, evaluation, and science network. Participant list is not treated as complete.",
    "focus_areas": [
      "evaluation",
      "testing_infrastructure",
      "international_dialogue",
      "dangerous_capability_evaluation",
      "standards"
    ],
    "tags": [
      "IN-AIMES",
      "international network",
      "AISI network naming",
      "measurement and evaluation"
    ],
    "source_ids": [
      "uk-aus-mou",
      "industry-au-seoul-declaration-2024"
    ],
    "confidence_level": "medium",
    "last_verified": "2026-06-12",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "A definitive official public participant roster was not located in the PR2G source set."
    ],
    "evidence_note_ids": [
      "ev-06-in-aimes-naming"
    ]
  },
  {
    "id": "aus-aisi",
    "name_en": "Australian AI Safety Institute",
    "acronym": "Australian AISI",
    "entity_type": "ai_safety_institute",
    "country": "Australia",
    "region": "Oceania",
    "summary": "APAC counterpart imported because the PR2G sources directly tie it to UK AISI through a formal 2026 MoU and UK Alignment Project coalition materials.",
    "focus_areas": [
      "frontier_ai_safety",
      "evaluation",
      "testing_infrastructure",
      "international_dialogue"
    ],
    "tags": [
      "Australia",
      "APAC comparator",
      "UK-Australia MoU",
      "Alignment Project"
    ],
    "source_ids": [
      "uk-aus-mou",
      "uk-alignment-project"
    ],
    "confidence_level": "high",
    "last_verified": "2026-06-06",
    "tier": 1,
    "needs_primary_source": false,
    "open_questions": [
      "A direct Australian official institute page would improve this profile beyond UK government sourcing."
    ],
    "evidence_note_ids": [
      "ev-05-uk-aus-mou",
      "ev-07-uk-alignment-project"
    ]
  },
  {
    "id": "oxford-martin-ai-governance-initiative",
    "name_en": "Oxford Martin AI Governance Initiative",
    "acronym": "OxMAGI",
    "entity_type": "research_group",
    "country": "United Kingdom",
    "region": "Europe",
    "summary": "Oxford governance initiative imported only for its directly supported IDAIS role, especially the London 2026 co-hosting claim and Beijing/series linkage.",
    "focus_areas": [
      "ai_governance",
      "international_dialogue",
      "technical_safety_research",
      "frontier_ai_safety",
      "dialogue_channel"
    ],
    "tags": [
      "Oxford Martin",
      "IDAIS",
      "London 2026 co-host",
      "dialogue channel"
    ],
    "source_ids": [
      "farai-idais-page",
      "idais-beijing"
    ],
    "confidence_level": "medium",
    "last_verified": "2026-06-06",
    "tier": 2,
    "needs_primary_source": false,
    "open_questions": [
      "A direct Oxford initiative page would strengthen institutional profile details beyond IDAIS pages."
    ],
    "evidence_note_ids": [
      "ev-14-idais-home",
      "ev-14-idais-beijing"
    ]
  }
]