{
  "id": "EM-AI-03",
  "name": "Оценка AI и доказательства безопасности",
  "domain": "AI",
  "kind": "subject",
  "wave": "W3",
  "scope": "План оценки, benchmark, запуск, результаты, оценка рисков и решение о допустимом применении. Метрика, dataset и evidence переиспользуются.",
  "candidate_types": [
    "EvaluationPlan",
    "EvaluationRun",
    "BenchmarkSpecification",
    "SafetyAssessment",
    "DeploymentDecision"
  ],
  "specific_questions": [
    "Как выявлять утечку benchmark в training?",
    "Как отделить измерение способности от допустимости применения?",
    "Как учитывать неопределённость и смену контекста?"
  ],
  "proposed_invariants": [
    "Результат закрепляет model и dataset версии",
    "Оценка не универсальна вне контекста",
    "Решение имеет ответственного и условия"
  ],
  "negative_case": "Один высокий benchmark score объявляет систему безопасной для всех задач.",
  "acceptance_scenario": "Смена датасета и контекста применения приводит к отдельной оценке и явному решению, а не к переносу прежнего одобрения.",
  "comparison_tracks": [
    "NIST AI RMF: контекст, оценка и управление риском",
    "PROV/SPDX: происхождение данных, артефактов и действий",
    "Практика model registry/evaluation tooling; сопоставить system, artifact, run и endpoint"
  ],
  "vercy_candidates": [
    {
      "model_id": "WM-AI-003",
      "mapping_status": "conceptual-candidate",
      "evidence_depth": "index-and-publication-metadata",
      "note": "Связь требует проверки полного семантического crosswalk; published не означает завершённую независимую экспертизу.",
      "runtime": {
        "id": "vr.wm-ai-003",
        "modelId": "WM-AI-003",
        "name": "AI Model Evaluation",
        "version": "0.3.0-research.1",
        "status": "published",
        "installable": true,
        "pageUrl": "https://ver.cy/models/wm-ai-003-ai-model-evaluation/",
        "specUrl": "https://ver.cy/models/wm-ai-003-ai-model-evaluation/spec.yaml",
        "agentsUrl": "https://ver.cy/models/wm-ai-003-ai-model-evaluation/AGENTS.md",
        "digest": "sha256:44d8e13c6124fd1b01f3b7481f61d0a119c33215084028cf384f52a2df8a7beb",
        "requires": [],
        "unresolvedRequires": []
      }
    },
    {
      "model_id": "WM-AI-009",
      "mapping_status": "conceptual-candidate",
      "evidence_depth": "index-and-publication-metadata",
      "note": "Связь требует проверки полного семантического crosswalk; published не означает завершённую независимую экспертизу.",
      "runtime": {
        "id": "vr.wm-ai-009",
        "modelId": "WM-AI-009",
        "name": "Evaluation Dataset / Benchmark",
        "version": "0.3.0-research.1",
        "status": "published",
        "installable": true,
        "pageUrl": "https://ver.cy/models/wm-ai-009-evaluation-dataset-benchmark/",
        "specUrl": "https://ver.cy/models/wm-ai-009-evaluation-dataset-benchmark/spec.yaml",
        "agentsUrl": "https://ver.cy/models/wm-ai-009-evaluation-dataset-benchmark/AGENTS.md",
        "digest": "sha256:1589441793fd64e51143d4f4b4a98fd0a27b7b101bb59fa02e7d66a974e02888",
        "requires": [],
        "unresolvedRequires": []
      }
    },
    {
      "model_id": "WM-AI-008",
      "mapping_status": "conceptual-candidate",
      "evidence_depth": "index-and-publication-metadata",
      "note": "Связь требует проверки полного семантического crosswalk; published не означает завершённую независимую экспертизу.",
      "runtime": {
        "id": "vr.wm-ai-008",
        "modelId": "WM-AI-008",
        "name": "AI Safety / Governance Assessment",
        "version": "0.3.0-research.1",
        "status": "published",
        "installable": true,
        "pageUrl": "https://ver.cy/models/wm-ai-008-ai-safety-governance-assessment/",
        "specUrl": "https://ver.cy/models/wm-ai-008-ai-safety-governance-assessment/spec.yaml",
        "agentsUrl": "https://ver.cy/models/wm-ai-008-ai-safety-governance-assessment/AGENTS.md",
        "digest": "sha256:d662be695593126b8a1b57322725a56d1aad0976f35e42c06166e6842c638a3c",
        "requires": [],
        "unresolvedRequires": []
      }
    },
    {
      "model_id": "WM-AI-010",
      "mapping_status": "conceptual-candidate",
      "evidence_depth": "index-and-publication-metadata",
      "note": "Связь требует проверки полного семантического crosswalk; published не означает завершённую независимую экспертизу.",
      "runtime": {
        "id": "vr.wm-ai-010",
        "modelId": "WM-AI-010",
        "name": "AI Incident Report",
        "version": "0.3.0-research.1",
        "status": "published",
        "installable": true,
        "pageUrl": "https://ver.cy/models/wm-ai-010-ai-incident-report/",
        "specUrl": "https://ver.cy/models/wm-ai-010-ai-incident-report/spec.yaml",
        "agentsUrl": "https://ver.cy/models/wm-ai-010-ai-incident-report/AGENTS.md",
        "digest": "sha256:1de749a3e9feaa2b68eba8153c08ac7a95e2901de465210a37089ea2b6265c72",
        "requires": [],
        "unresolvedRequires": []
      }
    }
  ],
  "candidate_properties_from_v1": [
    {
      "predecessor": "AI-06",
      "fields": [
        {
          "name": "protocol_version",
          "value_type": "text",
          "status": "candidate-not-normative"
        },
        {
          "name": "benchmark_version",
          "value_type": "text",
          "status": "candidate-not-normative"
        },
        {
          "name": "result",
          "value_type": "map",
          "status": "candidate-not-normative"
        },
        {
          "name": "limitations",
          "value_type": "text",
          "status": "candidate-not-normative"
        }
      ]
    },
    {
      "predecessor": "AI-08",
      "fields": [
        {
          "name": "assessment_scope",
          "value_type": "text",
          "status": "candidate-not-normative"
        },
        {
          "name": "risk_scenarios",
          "value_type": "text",
          "status": "candidate-not-normative"
        },
        {
          "name": "release_criteria",
          "value_type": "text",
          "status": "candidate-not-normative"
        },
        {
          "name": "monitoring_plan",
          "value_type": "text",
          "status": "candidate-not-normative"
        }
      ]
    }
  ],
  "suggested_owner": "Руководитель AI / исследований",
  "candidate_master_systems": "Model registry, experiment tracker, eval store",
  "related_research_contours": [
    "EM-AI-01",
    "EM-AI-02"
  ],
  "blocking_decisions": [
    "Установить границу и решение reuse/extend/new по действующим спецификациям.",
    "Подтвердить semantic crosswalk, права и source mastership.",
    "Выбрать immutable refs; провести проверки fixtures до заявления о публикационной готовности."
  ],
  "research_progress": {
    "sequence": 95,
    "id": "EM-AI-03",
    "status": "queued",
    "claude_status": "not-started",
    "grok_status": "not-started",
    "boundary_decision": "pending",
    "publication_urls": [],
    "remaining_scope": "Entire research brief pending",
    "target_model_ids": [
      "WM-AI-003",
      "WM-AI-009",
      "WM-AI-008",
      "WM-AI-010"
    ]
  },
  "brief_url": "https://ver.cy/enterprise/models/em-ai-03/"
}