{
  "meta": {
    "generatedAt": "2026-01-09T00:00:00.000Z",
    "count": 1,
    "permalink": "/standards/laws",
    "release": {
      "id": "2026.01",
      "label": "2026.01",
      "date": "2026-01-09",
      "permalink": "/api"
    }
  },
  "dependencies": [
    {
      "schema_version": "1.1.0",
      "dependency_id": "DEP-RISK-OPS-V2-2026-06",
      "system_ref": "agent-risk-ops-v2",
      "assessed_at": "2026-06-15T00:00:00Z",
      "dependents": [
        {
          "kind": "workflow",
          "name": "Account-takeover triage queue",
          "criticality": "critical"
        },
        {
          "kind": "role",
          "name": "Fraud analyst (tier 1)",
          "criticality": "high"
        },
        {
          "kind": "customer",
          "name": "Retail account holders",
          "criticality": "high"
        },
        {
          "kind": "downstream_software",
          "name": "Case management system hold sync",
          "criticality": "medium"
        }
      ],
      "substitution_cost": {
        "estimate": 26,
        "unit": "person-weeks",
        "basis": "Estimate to re-staff manual triage at current volume plus rules-engine fallback tuning."
      },
      "expertise_retained": {
        "roles": [
          "Fraud analyst (tier 2)",
          "Risk Steward"
        ],
        "last_exercised": "2026-05-20T00:00:00Z"
      },
      "alternatives": [
        {
          "name": "Manual triage by tier-2 analysts",
          "status": "degraded",
          "last_exercised": "2026-05-20T00:00:00Z"
        },
        {
          "name": "Legacy velocity rules engine",
          "status": "available",
          "last_exercised": "2026-05-20T00:00:00Z"
        }
      ],
      "last_withdrawal_rehearsal": {
        "at": "2026-05-20T00:00:00Z",
        "outcome": "partial",
        "notes": "Rules engine took over within 40 minutes; manual triage backlog reached 6 hours before clearing. Analyst headcount is the constraint."
      },
      "correction_latency": {
        "estimate_hours": 31,
        "basis": "Median time from challenge to hold release, measured by the latency tracker over Q2."
      },
      "reversibility": {
        "technical": {
          "feasible": true,
          "evidence": "holds.release reverses any hold; verified in rehearsal 2026-05-20."
        },
        "operational": {
          "feasible": true,
          "evidence": "Risk Steward released 100% of rehearsal holds within the 2-hour reach target."
        },
        "institutional": {
          "feasible": false,
          "evidence": "Withdrawal beyond 48 hours breaches the fraud-loss SLA with the card network; not survivable without the rules-engine fallback being fully staffed."
        }
      },
      "preserved_capacities": [
        {
          "capacity": "Tier-2 manual triage staffing at 30% of peak volume",
          "reason": "Keeps withdrawal survivable and preserves the ability to question the system's decisions.",
          "owner": "Fraud Operations lead",
          "last_exercised": "2026-05-20T00:00:00Z",
          "status": "degrading",
          "change_since_last_assessment": "consumed"
        },
        {
          "capacity": "Legacy rules engine kept deployable",
          "reason": "Substitution path that does not depend on the agent vendor.",
          "owner": "Platform engineering",
          "last_exercised": "2026-05-20T00:00:00Z",
          "status": "retained",
          "change_since_last_assessment": "unchanged"
        },
        {
          "capacity": "Ability to re-specify the hold rules for a replacement vendor without the incumbent's model documentation",
          "reason": "Substitution requires being able to state the requirement independently of the incumbent; without it the only available substitute is the incumbent's next version.",
          "owner": "Model Risk Review (second line)",
          "last_exercised": "2025-09-30T00:00:00Z",
          "status": "lost",
          "change_since_last_assessment": "consumed"
        }
      ],
      "independent_evaluation": {
        "executor": "Fraud Operations desk, which runs agent-risk-ops-v2 and owns the hold queue",
        "evaluator": "Model Risk Review (second line)",
        "independent": true,
        "evidence": "Q2 false-positive rate and release latency were recomputed by Model Risk Review from the raw hold log rather than accepted from the Fraud Operations dashboard; recomputation signed off 2026-06-12. Independence covers evaluation only: the capacity to re-specify the rules for a replacement vendor is recorded as lost, so the evaluator can say the system is wrong without being able to say what should replace it."
      },
      "exposure_score": {
        "dependency_depth": 3,
        "substitution_cost": 26,
        "correction_latency": 31,
        "score": 2418
      }
    }
  ]
}