{
  "model": {
    "id": "dttt-environmental",
    "name": "DTTT AI Environmental Impact Model",
    "version": "0.1",
    "release_date": "2026-03-07",
    "release_type": "pre-release",
    "status": "pre-release",
    "description": "A framework for measuring and communicating the environmental footprint of AI use in advisory and destination marketing work. Version 0.1 establishes the energy intensity categorisation and disclosure format. A grading scale will be introduced at v0.5 following methodology review.",
    "url": "https://thedttt.com/ai-transparency",
    "data_confidence": "Estimated",
    "disclaimer": "Energy and carbon estimates for AI tool usage are contested and the underlying science is developing rapidly. Scores in this model are indicative and based on published research. They are not audited figures. DTTT will update the methodology sources with each version."
  },
  "energy_intensity_categories": [
    {
      "category": "Low",
      "description": "Local inference, small open-source models, lightweight AI-assisted tools running on device or on modest infrastructure.",
      "examples": ["On-device spell-check and grammar tools", "Small local language models", "Lightweight AI writing assistants"]
    },
    {
      "category": "Medium",
      "description": "Standard cloud-based AI tools at typical usage levels. Most general-purpose AI assistants fall here for standard text-based tasks.",
      "examples": ["Standard conversational AI for research and drafting", "AI-assisted image editing", "Automated data processing pipelines"]
    },
    {
      "category": "High",
      "description": "Large frontier model usage, image generation, video generation and extended compute tasks. Energy intensity significantly higher per query.",
      "examples": ["Large frontier models for extended research sessions", "AI image generation", "AI video generation", "Complex data modelling and simulation"]
    }
  ],
  "disclosure_format": {
    "fields": [
      {
        "field": "energy_category",
        "label": "Energy category",
        "type": "single_select",
        "options": ["Low", "Medium", "High"]
      },
      {
        "field": "carbon_band",
        "label": "Estimated carbon equivalent",
        "type": "single_select",
        "options": ["Negligible", "Low", "Moderate", "Significant"],
        "note": "Indicative only. Based on published energy intensity research."
      },
      {
        "field": "mitigation",
        "label": "Mitigation actions",
        "type": "multi_select",
        "options": [
          "Model selection (chose lower-intensity model where possible)",
          "Usage reduction (limited queries to what was necessary)",
          "Carbon offsetting",
          "Renewable energy provider",
          "None taken"
        ]
      },
      {
        "field": "disclosure_note",
        "label": "Disclosure note for deliverable footer",
        "type": "text",
        "template": "AI tools used in producing this work were categorised as [energy_category] energy intensity. Estimated carbon impact: [carbon_band]. [mitigation note if applicable]."
      }
    ]
  },
  "grading_scale": {
    "status": "deferred",
    "target_version": "0.5",
    "note": "A grading scale consistent with the A–E structure of the AI Transparency Model will be introduced at v0.5, following committee review of the methodology sources and carbon band definitions."
  },
  "methodology_sources": [
    {
      "source": "IEA",
      "title": "Electricity 2024: Analysis and Forecast to 2026",
      "year": 2024,
      "relevance": "Data centre electricity consumption projections including AI workloads"
    },
    {
      "source": "Goldman Sachs",
      "title": "AI is Poised to Drive 160% Increase in Data Center Power Demand",
      "year": 2024,
      "relevance": "AI power demand projections"
    }
  ],
  "governance": {
    "licence": "Creative Commons Attribution 4.0 International",
    "contact": "hello@thedttt.com",
    "committee": true,
    "versioning_policy": "Pre-release versions (0.x) are published openly but clearly labelled as in development. v0.5 introduces the grading scale. v1.0 marks first stable release with verified methodology sources."
  },
  "changelog": [
    {
      "version": "0.1",
      "date": "2026-03-07",
      "type": "pre-release",
      "summary": "Initial pre-release. Establishes scope, energy intensity categories and disclosure format. Grading scale deferred to v0.5.",
      "changes": []
    }
  ]
}