GIRAI 2026 launches 38-indicator index to score responsible AI governance
135 researchers across countries measured AI policy enforcement, not just promises.
The Global Index on Responsible AI (GIRAI), led by authors including Fola Adeleke, Rachel Adams, and Ayantola Alayande, published its 2nd Edition conceptual framework and methodology on arXiv. Building on the 1st Edition, this version sharpens the divide between whether AI governance frameworks exist and whether they are actually implemented. The new model expands three thematic areas into five dimensions—Inclusion and Diversity, Ethics and Sustainability, Labour and Skills, Trust and Safety, and Use of AI in Public Service—backed by 38 indicators total.
The methodology assigns 17 indicators to AI Policy, 5 to Civil Society Organization (CSO) Engagement, and 15 to Enabling Conditions, with pillar weights of 60%, 10%, and 30% respectively. An independent statistical pre-audit tested framework coherence, and a separate government Use of Unacceptable Risk AI (URAI) indicator acts as an accountability penalty. Data was collected via a structured global survey by 135 country-level researchers, combined with secondary datasets. All data is normalized to a 100-point scale for systematic cross-national comparison. This gives policymakers, civil society, and AI developers a clear view of where commitments translate into enforceable protections—and where gaps remain.
- GIRAI 2026 expands to 38 indicators across 5 dimensions, up from the 1st Edition's 3 thematic areas
- Scoring weights: 60% AI policy, 30% enabling conditions, 10% CSO engagement, plus a URAI deduction penalty
- Data gathered by 135 country-level researchers via structured surveys and secondary datasets
Why It Matters
A standardized global metric lets governments and firms benchmark AI accountability and target enforcement gaps.