EU AI Act : obligations haut risque à partir d'août 2026Vérifiez votre conformité →

Sources & Methodology

AI Pulse is built on open, trusted data from three complementary sources, processed with the TokenFlop methodology.

Construit sur des données ouvertes et fiables

fruggr

Base de lois IA Fruggr

133 lois et réglementations suivies dans 46 juridictions. Maintenue par notre équipe gouvernance.

MIT AIID

MIT AI Risk Repository

1 700+ risques IA catalogués sur 7 domaines avec taxonomie causale.

OECD

Observatoire des politiques IA de l'OCDE

Indices d'adoption nationale de l'IA et indicateurs politiques dans 80+ juridictions.

TokenFlop Methodology

TokenFlop is an open, physics-based computational model developed by Digital4Better that estimates the environmental footprint of AI systems. It converts FLOPs into GPU hours, then into energy consumption, and finally into CO₂ equivalent — factoring in hardware type, datacenter PUE, energy mix by region, and manufacturing footprint.

The methodology follows ISO 14040 / ITU L.1410 standards and covers training, fine-tuning and inference across text, image, audio and video modalities.

Read the full methodology

Credits & Acknowledgements

AI Pulse is maintained by the Digital4Better / fruggr team. The AI law database is curated monthly by our governance experts. Risk data comes from the MIT AI Risk Repository (Slattery et al., 2024, CC BY 4.0). Incident data is sourced from the OECD AI Incidents Monitor. Adoption indices from the OECD AI Policy Observatory.