{
  "notice": "Illustrative mapping from the AI Governance Engineer Body of Knowledge v0.5.0 (not a claim of conformity)",
  "version": "0.5.0",
  "license": "CC BY 4.0",
  "licenseUrl": "https://creativecommons.org/licenses/by/4.0/",
  "schemaVersion": 1,
  "schema": "https://aigovernanceengineer.com/api/v1/schemas/control.json",
  "self": "https://aigovernanceengineer.com/api/v1/controls/aige-ctl-assure-002.json",
  "source": "https://aigovernanceengineer.com/controls/assurance-and-evidence#aige-ctl-assure-002",
  "citation": {
    "title": "AI Governance Engineering: The Thesis & Body of Knowledge",
    "authors": [
      "Jorge García Aibar"
    ],
    "parentDoi": "https://doi.org/10.5281/zenodo.22956197",
    "conceptDoi": "https://doi.org/10.5281/zenodo.22857084"
  },
  "control": {
    "id": "AIGE-CTL-ASSURE-002",
    "profile": "assurance-and-evidence",
    "url": "https://aigovernanceengineer.com/controls/assurance-and-evidence#aige-ctl-assure-002",
    "json": "https://aigovernanceengineer.com/api/v1/controls/aige-ctl-assure-002.json",
    "title": "Release Blocked Below the Eval Threshold",
    "version": "0.1",
    "status": "draft",
    "reviewerStatus": "open",
    "depth": "derived",
    "objective": "A model or agent ships only after a versioned eval suite, with at least one capability and one adversarial eval, passes in the pipeline above a documented threshold that traces to a named failure mode or obligation, and every run leaves a structured result filed against the registry entry of the version tested.",
    "failureModes": [
      "A model or agent is retrained, re-prompted or given a new tool and ships with no eval run for its version.",
      "A result below the threshold does not fail the pipeline, so the release ships and a finding is filed instead.",
      "A threshold traces to no named failure mode or obligation.",
      "A result exists only as a pasted score or a slide, not as a structured record (suite id, model version, score, threshold, result, timestamp) filed against the registry entry."
    ],
    "scope": "Models and agents that change (retrained, re-prompted or given a new tool) and ship through a pipeline that already runs functional tests. The trajectory evals of an agent are AIGE-CTL-AGENT-013, and the checks that a result is valid enough to gate a release are AIGE-CTL-EVAL-009.",
    "enforcementPoints": [
      "pre_merge"
    ],
    "verification": [],
    "evidence": [
      {
        "artefact": "Eval result of each run: suite id, model version, score, threshold, pass or fail and timestamp, filed against the registry entry",
        "schemaId": "eval-result",
        "schema": "https://aigovernanceengineer.com/schemas/eval-result.v1.json",
        "layer": 3
      }
    ],
    "failureResponse": {
      "effect": "deny",
      "text": "A result below the threshold fails the pipeline, and the release does not ship until it is fixed."
    },
    "layer": 3,
    "secondaryLayers": [],
    "patterns": [
      {
        "slug": "eval-gate-in-ci",
        "title": "Eval Gate in CI",
        "url": "https://aigovernanceengineer.com/patterns/eval-gate-in-ci"
      },
      {
        "slug": "adversarial-red-team-suite",
        "title": "Adversarial Red-Team Suite",
        "url": "https://aigovernanceengineer.com/patterns/adversarial-red-team-suite"
      }
    ],
    "seeds": [],
    "derivedFrom": [
      {
        "kind": "pattern",
        "ref": "eval-gate-in-ci",
        "url": "https://aigovernanceengineer.com/patterns/eval-gate-in-ci"
      },
      {
        "kind": "schema",
        "ref": "eval-result",
        "url": "https://aigovernanceengineer.com/resources/templates#schema-eval-result"
      },
      {
        "kind": "chapter",
        "ref": "governing-development",
        "url": "https://aigovernanceengineer.com/bok/governing-development"
      }
    ],
    "mappings": {
      "obligations": [
        {
          "id": "AIGE-OBL-EUAIA-ART15",
          "name": "EU AI Act Art. 15 accuracy, robustness and cybersecurity",
          "url": "https://aigovernanceengineer.com/obligations/aige-obl-euaia-art15"
        },
        {
          "id": "AIGE-OBL-EUAIA-ART55",
          "name": "EU AI Act Art. 55 GPAI models with systemic risk",
          "url": "https://aigovernanceengineer.com/obligations/aige-obl-euaia-art55"
        },
        {
          "id": "AIGE-OBL-NISTRMF-MEASURE",
          "name": "MEASURE",
          "url": "https://aigovernanceengineer.com/obligations/aige-obl-nistrmf-measure"
        },
        {
          "id": "AIGE-OBL-OWASP-AGENTIC",
          "name": "Top 10 for Agentic Applications 2026",
          "url": "https://aigovernanceengineer.com/obligations/aige-obl-owasp-agentic"
        }
      ],
      "iso42001": [
        {
          "id": "A.6.2.4",
          "title": "AI system verification and validation"
        }
      ],
      "nistAiRmf": [
        {
          "id": "MEASURE 2.3",
          "title": "Performance or assurance criteria measured for deployment-like conditions"
        }
      ],
      "owasp": [
        {
          "id": "asi01",
          "externalId": "ASI01",
          "name": "Agent Goal Hijack",
          "url": "https://aigovernanceengineer.com/resources/threats#threat-asi01"
        },
        {
          "id": "asi02",
          "externalId": "ASI02",
          "name": "Tool Misuse and Exploitation",
          "url": "https://aigovernanceengineer.com/resources/threats#threat-asi02"
        }
      ],
      "atlas": [],
      "aiuc1": [
        "C002"
      ],
      "csaAicm": [],
      "other": []
    },
    "references": [
      {
        "n": 2,
        "title": "Pattern: Eval Gate in CI",
        "text": "Pattern: Eval Gate in CI (AI Governance Engineering Body of Knowledge v0.5.0, chapter 05 pattern catalogue). AI Governance Engineer (Jorge García Aibar). 2026-09.",
        "url": "https://aigovernanceengineer.com/patterns/eval-gate-in-ci",
        "verified": "primary"
      },
      {
        "n": 6,
        "title": "Governing AI development",
        "text": "Governing AI development (AI Governance Engineering Body of Knowledge v0.5.0, chapter 14, section \"Statistical validity of evals\"). AI Governance Engineer (Jorge García Aibar). 2026-09.",
        "url": "https://aigovernanceengineer.com/bok/governing-development#statistical-validity-of-evals",
        "verified": "primary"
      },
      {
        "n": 7,
        "title": "Governing AI development",
        "text": "Governing AI development (AI Governance Engineering Body of Knowledge v0.5.0, chapter 14, section \"Reproducibility and linked versioning\"). AI Governance Engineer (Jorge García Aibar). 2026-09.",
        "url": "https://aigovernanceengineer.com/bok/governing-development#reproducibility-and-linked-versioning",
        "verified": "primary"
      },
      {
        "n": 8,
        "title": "OWASP Top 10 for Agentic Applications for 2026",
        "text": "OWASP Top 10 for Agentic Applications for 2026 (ASI01 Agent Goal Hijack to ASI10 Rogue Agents). OWASP GenAI Security Project. 2025-12-09.",
        "url": "https://genai.owasp.org/resource/owasp-top-10-for-agentic-applications-for-2026/",
        "verified": "primary"
      },
      {
        "n": 3,
        "title": "Regulation (EU) 2024/1689 (AI Act), consolidated text of 2026-07-27 as amended by Regulation (EU) 2026/1744",
        "text": "Regulation (EU) 2024/1689 (AI Act), consolidated text of 2026-07-27 as amended by Regulation (EU) 2026/1744 (the articles each control maps to, as chapters 14 and 18 restate them). Publications Office of the EU (EUR-Lex). 2026-07-27.",
        "url": "https://eur-lex.europa.eu/eli/reg/2024/1689/2026-07-27/eng",
        "verified": "primary"
      },
      {
        "n": 4,
        "title": "Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1",
        "text": "Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1 (MEASURE and MANAGE subcategories cited by id, mapped only where the official text matches the control). NIST. 2023-01-26.",
        "url": "https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.100-1.pdf",
        "verified": "primary"
      },
      {
        "n": 5,
        "title": "ISO/IEC 42001:2023, AI management systems",
        "text": "ISO/IEC 42001:2023, AI management systems (Annex A control ids and clause numbers cited by number and short title only; the text of the standard was not opened). ISO/IEC. 2023-12.",
        "url": "https://www.iso.org/standard/81230.html",
        "verified": "secondary"
      },
      {
        "n": 9,
        "title": "AIUC-1 requirements",
        "text": "AIUC-1 requirements (public requirement index, A001 to F002, each requirement on its own page (E007 and E014 marked retired); AIUC-1 is a standard of the Artificial Intelligence Underwriting Company; this site is not affiliated with AIUC, and a mapping here is not an AIUC-1 certificate or audit). Artificial Intelligence Underwriting Company. 2026-09-24.",
        "url": "https://standard.aiuc-1.com/llms.txt",
        "verified": "primary"
      }
    ],
    "implementationNotes": [
      "Version the suite alongside the model and run it in CI (for example with Inspect, promptfoo, Garak or Giskard; illustrative); the security suite is the Adversarial Red-Team Suite.",
      "Size the suite from the threshold, not from the time available: chapter 14 shows a 0.96 pass rate on 200 cases with a 95% interval of about 0.933 to 0.987, which a 0.95 threshold sits inside, so the gate cannot tell a pass from a fail.",
      "Link the records both ways: model version to training record to eval results to release tag to the risk approvals that let it ship."
    ],
    "openQuestions": [
      "Verification procedure to be specified: the derivation adds no check its source material does not state; requires technical review.",
      "How should the gate treat a suite whose repeated runs straddle the threshold: rerun, enlarge the sample or block, given that the pattern warns against flaky gates?"
    ],
    "observation": null,
    "observationSchema": "https://aigovernanceengineer.com/schemas/control-observation.v1.json",
    "examples": []
  }
}
