ISO 14064 Compliance

Guide for greenhouse gas reporting under ISO 14064-1:2018.

What is ISO 14064?

ISO 14064 is the international standard for quantifying and reporting greenhouse gas emissions. It provides detailed requirements for organizational-level GHG inventories.

ISO 14064 has three parts: - Part 1: Organization-level quantification and reporting - Part 2: Project-level quantification and reporting - Part 3: Verification and validation

The Green module supports Part 1 requirements.

Key Requirements

Organizational Boundary

Define which operations are included:

Organizational boundary: Operational control
Included: AI inference API usage by [Organization]
Excluded: Direct operations (reported separately)

Operational Boundary

Specify emission sources:

Scope 3, Category 1: Purchased goods and services
Source: AI inference API calls
Activity: Token processing (input, output, cache)

Base Year

Establish reference point for tracking:

openclaw green targets:add \
  --name "Base Year Inventory" \
  --base-year 2025 \
  --target-year 2030 \
  --reduction 0

The base year inventory captures emissions at a fixed point.

Uncertainty Quantification

ISO 14064 requires uncertainty assessment. OpenClaw calculates uncertainty from confidence scores:

Confidence Uncertainty Range
≥70% ±15%
≥50% ±30%
≥30% ±50%
<30% ±100%

View uncertainty:

openclaw green intensity

Output includes uncertainty range.

Export Format

ISO 14064 doesn't prescribe a specific format. Use TCFD export with additional metadata:

openclaw green export --format tcfd --period 2025 --baseline 2024

Add ISO 14064-specific fields manually:

{
  "standard": "ISO 14064-1:2018",
  "reportingPeriod": {
    "start": "2025-01-01",
    "end": "2025-12-31"
  },
  "organizationalBoundary": {
    "approach": "operational_control",
    "entities": ["[Organization name]"]
  },
  "operationalBoundary": {
    "scope3": {
      "category1": {
        "included": true,
        "sources": ["AI inference API calls"]
      }
    }
  },
  "quantificationMethodology": {
    "method": "calculation",
    "activityData": "API token counts",
    "emissionFactors": "Academic research (Lacoste et al. 2019)",
    "globalWarmingPotentials": "IPCC AR6"
  },
  "uncertainty": {
    "method": "confidence-based",
    "overallUncertainty_percent": 30
  },
  "emissions": {
    "scope3Cat1_tCO2eq": 0.01245,
    "uncertainty_tCO2eq": 0.00374
  }
}

Uncertainty Methodology

Calculation

Uncertainty (%) = f(confidence)

where:
  confidence ≥ 0.7 → ±15%
  confidence ≥ 0.5 → ±30%
  confidence ≥ 0.3 → ±50%
  confidence < 0.3 → ±100%

Reporting

Express emissions with uncertainty:

12.45 kg CO₂eq ± 3.74 kg (±30%)

Or as range:

8.72 - 16.19 kg CO₂eq (95% confidence interval)

Documentation Requirements

ISO 14064 requires extensive documentation:

1. GHG Management Plan

## AI Emissions Management

### Scope
AI inference emissions from API usage.

### Responsibility
[Role] responsible for data collection and reporting.

### Frequency
Quarterly calculation, annual reporting.

### Data Sources
- API usage logs (token counts)
- Provider invoices (validation)
- OpenClaw Green module (calculation)

2. Quantification Methodology

## Quantification Methodology

### Activity Data
Token counts (input, output, cache) from API responses.

### Emission Factors
Per-model factors from academic research:
- Lacoste et al. (2019)
- Patterson et al. (2022)
- CodeCarbon project

### Calculation
CO₂ = Σ(tokens × factor_per_token)

### Global Warming Potentials
IPCC AR6 (CO₂ = 1, no other gases for inference)

3. Uncertainty Assessment

## Uncertainty Assessment

### Sources of Uncertainty
1. Emission factors (±25%) - no supplier data
2. Activity data (±5%) - API logs accurate
3. Model mapping (±10%) - model aliases

### Combined Uncertainty
±30% (root sum of squares)

### Limitations
- No supplier-specific data
- Data center PUE assumed
- Grid carbon intensity averaged

Verification Readiness

For third-party verification under ISO 14064-3:

Evidence to Provide

  1. Raw data: openclaw green export --format json
  2. Summary: openclaw green export --format tcfd
  3. Methodology: Link to this documentation
  4. Factor sources: Academic papers
  5. Activity logs: API usage records

Verification Statement

The GHG inventory for AI inference emissions has been verified
to a limited/reasonable level of assurance in accordance with
ISO 14064-3:2019.

Scope: Scope 3, Category 1 emissions from AI inference API usage
Period: January 1 - December 31, 2025
Emissions: 12.45 kg CO₂eq ± 30%

Continuous Improvement

ISO 14064 emphasizes continuous improvement:

  1. Annual review of emission factors
  2. Engage providers for better data
  3. Reduce uncertainty through measurement
  4. Set targets for emission reduction
  5. Track progress against base year

Example Report Structure

# GHG Inventory Report - AI Emissions
## ISO 14064-1:2018 Compliant

### 1. Organization Description
[Description of organization and AI usage]

### 2. Organizational Boundary
Operational control approach.

### 3. Operational Boundary
Scope 3, Category 1: AI inference API calls

### 4. Quantification Methodology
[Detailed methodology description]

### 5. GHG Emissions
| Category | Emissions | Uncertainty |
|----------|-----------|-------------|
| Scope 3 Cat 1 | 12.45 kg CO₂eq | ±30% |

### 6. Base Year
2025: 12.45 kg CO₂eq

### 7. Uncertainty Assessment
[Detailed uncertainty analysis]

### 8. Management Responsibility
[Roles and responsibilities]

### 9. Verification Statement
[If applicable]
  • GHG Protocol — Compatible methodology
  • TCFD — Disclosure framework
  • SBTi — Target-setting aligned with ISO 14064