{
  "name": "AI Agent Governance Maturity Model",
  "description": "A five-level maturity model for assessing how well an organization governs its autonomous AI agents.",
  "url": "https://gianlucabusato.com/datasets/ai-agent-governance-maturity-model/",
  "creator": {
    "name": "Gianluca Busato",
    "id": "https://gianlucabusato.com/#person"
  },
  "dateModified": "2026-07-15",
  "inLanguage": "en",
  "columns": [
    "Level",
    "Stage",
    "Characteristics"
  ],
  "records": [
    {
      "Level": "1",
      "Stage": "Ad hoc",
      "Characteristics": "Agents deployed without identity, policy or audit; behavior is opaque."
    },
    {
      "Level": "2",
      "Stage": "Logged",
      "Characteristics": "Actions are recorded after the fact, but not constrained in real time."
    },
    {
      "Level": "3",
      "Stage": "Controlled",
      "Characteristics": "Scoped permissions and basic policy enforcement; human approval for high-impact actions."
    },
    {
      "Level": "4",
      "Stage": "Governed",
      "Characteristics": "Multi-layer governance: verifiable identity, deterministic policy, mediated execution, immutable audit."
    },
    {
      "Level": "5",
      "Stage": "Autonomous & accountable",
      "Characteristics": "Fleets of orchestrated agents operate autonomously within deterministic guardrails, fully transparent and reversible."
    }
  ]
}