{
  "code": null,
  "collection": "architecture",
  "href": "https://banes-lab.com/ontology#architecture-model-safety",
  "id": "model-safety",
  "kind": "quality-attribute",
  "layer": {
    "href": "https://banes-lab.com/ontology/schema#layer-correctness-core",
    "json": "https://banes-lab.com/json/records/layer/correctness-core",
    "label": "Correctness Core",
    "markdown": "https://banes-lab.com/records/layer/correctness-core.md",
    "ref": "layer:correctness-core"
  },
  "name": "Model Safety",
  "ref": "architecture:model-safety",
  "relations": [
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-evaluation",
          "json": "https://banes-lab.com/json/records/lexicon/evaluation",
          "label": "Evaluation",
          "markdown": "https://banes-lab.com/records/lexicon/evaluation.md",
          "ref": "lexicon:evaluation"
        },
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-guardrails",
          "json": "https://banes-lab.com/json/records/lexicon/guardrails",
          "label": "Guardrails",
          "markdown": "https://banes-lab.com/records/lexicon/guardrails.md",
          "ref": "lexicon:guardrails"
        },
        {
          "href": "https://banes-lab.com/ontology#architecture-monitoring",
          "json": "https://banes-lab.com/json/records/architecture/monitoring",
          "label": "Monitoring",
          "markdown": "https://banes-lab.com/records/architecture/monitoring.md",
          "ref": "architecture:monitoring"
        }
      ],
      "relation": "requires"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology#architecture-model-governance",
          "json": "https://banes-lab.com/json/records/architecture/model-governance",
          "label": "Model Governance",
          "markdown": "https://banes-lab.com/records/architecture/model-governance.md",
          "ref": "architecture:model-governance"
        },
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-security",
          "json": "https://banes-lab.com/json/records/lexicon/security",
          "label": "Security",
          "markdown": "https://banes-lab.com/records/lexicon/security.md",
          "ref": "lexicon:security"
        }
      ],
      "relation": "reinforces"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-safe-model-deployment",
          "json": "https://banes-lab.com/json/records/lexicon/safe-model-deployment",
          "label": "Safe Model Deployment",
          "markdown": "https://banes-lab.com/records/lexicon/safe-model-deployment.md",
          "ref": "lexicon:safe-model-deployment"
        }
      ],
      "relation": "enables"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-unguarded-model-autonomy",
          "json": "https://banes-lab.com/json/records/lexicon/unguarded-model-autonomy",
          "label": "Unguarded Model Autonomy",
          "markdown": "https://banes-lab.com/records/lexicon/unguarded-model-autonomy.md",
          "ref": "lexicon:unguarded-model-autonomy"
        }
      ],
      "relation": "conflicts-with"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-capability-utility",
          "json": "https://banes-lab.com/json/records/lexicon/capability-utility",
          "label": "Capability/Utility",
          "markdown": "https://banes-lab.com/records/lexicon/capability-utility.md",
          "ref": "lexicon:capability-utility"
        }
      ],
      "relation": "tensions-with"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology/schema#tension-capability-utility-model-safety",
          "json": "https://banes-lab.com/json/records/tension/capability-utility-model-safety",
          "label": "Model Safety / Capability/Utility",
          "markdown": "https://banes-lab.com/records/tension/capability-utility-model-safety.md",
          "ref": "tension:capability-utility-model-safety"
        }
      ],
      "relation": "tensions"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-unguarded-model-autonomy",
          "json": "https://banes-lab.com/json/records/lexicon/unguarded-model-autonomy",
          "label": "Unguarded Model Autonomy",
          "markdown": "https://banes-lab.com/records/lexicon/unguarded-model-autonomy.md",
          "ref": "lexicon:unguarded-model-autonomy"
        }
      ],
      "relation": "violated-by"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-guardrails",
          "json": "https://banes-lab.com/json/records/lexicon/guardrails",
          "label": "Guardrails",
          "markdown": "https://banes-lab.com/records/lexicon/guardrails.md",
          "ref": "lexicon:guardrails"
        },
        {
          "href": "https://banes-lab.com/ontology#architecture-code-review",
          "json": "https://banes-lab.com/json/records/architecture/code-review",
          "label": "Code Review",
          "markdown": "https://banes-lab.com/records/architecture/code-review.md",
          "ref": "architecture:code-review"
        },
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-evaluation-suite",
          "json": "https://banes-lab.com/json/records/lexicon/evaluation-suite",
          "label": "Evaluation Suite",
          "markdown": "https://banes-lab.com/records/lexicon/evaluation-suite.md",
          "ref": "lexicon:evaluation-suite"
        }
      ],
      "relation": "refactored-by"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology/schema#vocabulary-severity-contextual",
          "json": "https://banes-lab.com/json/records/vocabulary/severity-contextual",
          "label": "contextual",
          "markdown": "https://banes-lab.com/records/vocabulary/severity-contextual.md",
          "ref": "vocabulary:severity-contextual"
        }
      ],
      "relation": "severity"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology#architecture-category-model-architecture",
          "json": "https://banes-lab.com/json/records/architecture-category/model-architecture",
          "label": "Model Architecture",
          "markdown": "https://banes-lab.com/records/architecture-category/model-architecture.md",
          "ref": "architecture-category:model-architecture"
        }
      ],
      "relation": "category"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology#architecture-artificial-intelligence-architecture",
          "json": "https://banes-lab.com/json/records/architecture/artificial-intelligence-architecture",
          "label": "Artificial Intelligence Architecture",
          "markdown": "https://banes-lab.com/records/architecture/artificial-intelligence-architecture.md",
          "ref": "architecture:artificial-intelligence-architecture"
        },
        {
          "href": "https://banes-lab.com/ontology#architecture-model-evaluation",
          "json": "https://banes-lab.com/json/records/architecture/model-evaluation",
          "label": "Model Evaluation",
          "markdown": "https://banes-lab.com/records/architecture/model-evaluation.md",
          "ref": "architecture:model-evaluation"
        },
        {
          "href": "https://banes-lab.com/ontology#architecture-agentic-architecture",
          "json": "https://banes-lab.com/json/records/architecture/agentic-architecture",
          "label": "Agentic Architecture",
          "markdown": "https://banes-lab.com/records/architecture/agentic-architecture.md",
          "ref": "architecture:agentic-architecture"
        }
      ],
      "relation": "reinforced-by"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/software-architecture/scale#systems-built-around-a-model",
          "json": "https://banes-lab.com/json/software-architecture/scale/systems-built-around-a-model",
          "label": "Systems built around a model",
          "markdown": "https://banes-lab.com/software-architecture/scale/systems-built-around-a-model.md",
          "ref": "chapter:/software-architecture/scale#systems-built-around-a-model"
        },
        {
          "href": "https://banes-lab.com/ontology#architecture-category-correctness-determinism-verification",
          "json": "https://banes-lab.com/json/ontology/principles/architecture-category-correctness-determinism-verification",
          "label": "Correctness / Determinism / Verification",
          "markdown": "https://banes-lab.com/ontology/principles/architecture-category-correctness-determinism-verification.md",
          "ref": "chapter:/ontology#architecture-category-correctness-determinism-verification"
        },
        {
          "href": "https://banes-lab.com/ontology#architecture-category-model-architecture",
          "json": "https://banes-lab.com/json/ontology/principles/architecture-category-model-architecture",
          "label": "Model Architecture",
          "markdown": "https://banes-lab.com/ontology/principles/architecture-category-model-architecture.md",
          "ref": "chapter:/ontology#architecture-category-model-architecture"
        },
        {
          "href": "https://banes-lab.com/ontology#architecture-category-observability-auditability-traceability",
          "json": "https://banes-lab.com/json/ontology/principles/architecture-category-observability-auditability-traceability",
          "label": "Observability / Auditability / Traceability",
          "markdown": "https://banes-lab.com/ontology/principles/architecture-category-observability-auditability-traceability.md",
          "ref": "chapter:/ontology#architecture-category-observability-auditability-traceability"
        },
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-category-core-vocabulary",
          "json": "https://banes-lab.com/json/ontology/lexicon/lexicon-category-core-vocabulary",
          "label": "Core Vocabulary",
          "markdown": "https://banes-lab.com/ontology/lexicon/lexicon-category-core-vocabulary.md",
          "ref": "chapter:/ontology/lexicon#lexicon-category-core-vocabulary"
        },
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-category-model-architecture",
          "json": "https://banes-lab.com/json/ontology/lexicon/lexicon-category-model-architecture",
          "label": "Model Architecture",
          "markdown": "https://banes-lab.com/ontology/lexicon/lexicon-category-model-architecture.md",
          "ref": "chapter:/ontology/lexicon#lexicon-category-model-architecture"
        },
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-category-quality-attributes",
          "json": "https://banes-lab.com/json/ontology/lexicon/lexicon-category-quality-attributes",
          "label": "Quality Attributes",
          "markdown": "https://banes-lab.com/ontology/lexicon/lexicon-category-quality-attributes.md",
          "ref": "chapter:/ontology/lexicon#lexicon-category-quality-attributes"
        },
        {
          "href": "https://banes-lab.com/ontology/schema#the-vocabulary-severity",
          "json": "https://banes-lab.com/json/ontology/schema/the-vocabulary-severity",
          "label": "Severity levels",
          "markdown": "https://banes-lab.com/ontology/schema/the-vocabulary-severity.md",
          "ref": "chapter:/ontology/schema#the-vocabulary-severity"
        },
        {
          "href": "https://banes-lab.com/ontology/schema#the-resolutions",
          "json": "https://banes-lab.com/json/ontology/schema/the-resolutions",
          "label": "The resolutions",
          "markdown": "https://banes-lab.com/ontology/schema/the-resolutions.md",
          "ref": "chapter:/ontology/schema#the-resolutions"
        }
      ],
      "relation": "linked-from"
    }
  ],
  "summary": "The degree to which a model-backed system prevents harmful inputs, outputs and actions through evaluation, guardrails and monitoring.",
  "siblings": {
    "next": {
      "href": "https://banes-lab.com/ontology#architecture-prompt-engineering",
      "json": "https://banes-lab.com/json/records/architecture/prompt-engineering",
      "label": "Prompt Engineering",
      "markdown": "https://banes-lab.com/records/architecture/prompt-engineering.md",
      "ref": "architecture:prompt-engineering"
    },
    "previous": {
      "href": "https://banes-lab.com/ontology#architecture-explainability",
      "json": "https://banes-lab.com/json/records/architecture/explainability",
      "label": "Explainability",
      "markdown": "https://banes-lab.com/records/architecture/explainability.md",
      "ref": "architecture:explainability"
    }
  },
  "up": {
    "href": null,
    "json": "https://banes-lab.com/json/api/records/architecture",
    "label": "Architecture principles",
    "markdown": "https://banes-lab.com/api/records/architecture.md",
    "ref": "api:records/architecture"
  },
  "aliases": [],
  "exemplar": {
    "after": "const input = await safety.validateInput(userPrompt);\nconst draft = await model.generate(input);\nconst checked = await safety.validateOutput(draft, { policy: \"foo-assistant-v2\" });\nif (!checked.allowed) return safeRefusal(checked.reasons);\nreturn checked.output;",
    "before": "return model.generate(userPrompt);",
    "lang": "ts",
    "medium": "code"
  },
  "formedBy": null,
  "scope": [
    "model-backed system",
    "model",
    "application"
  ],
  "severity": "contextual"
}