03/09/2026
From Open Data to AI-ready Data report is now online!
Published by the 開放文化基金會
Open Culture Foundation in collaboration with us, this report examines international best practices for transforming government open data to AI-Read, making it processable by AI tools.
Government open data must evolve into AI-ready open data to inspire innovation. Drawing on international examples, the report offers practical guidance for this transition. It explores what we can do in terms of data formats, metadata, licensing, privacy protection, cross-agency coordination and through public–private collaboration—showing how governments and civil society can work together to achieve it.
The report features the following international case studies:
🇪🇸 Spain | ImpulsaDATA
A programme that helps public institutions navigate overlapping regulations by accompanying public servants to walk through the process. Its diagnose–gap–act methodology enables agencies to identify high-value datasets that can be safely released.
🇩🇪 Germany | The German Commons
This project consolidates 41 clearly licensed German-language data sources and provides an automated code repository covering data cleaning, personally identifiable information removal, and other workflow steps for developing fully open German large language models.
🇩🇪 Germany | Parla
Jointly developed by the Senate Chancellery of Berlin and CityLAB Berlin, Parla extracts more than 11,000 public parliamentary documents and uses a generative AI chatbot to improve administrative searches and responses to citizens’ enquiries.
🇰🇭 Cambodia | Open Development Cambodia (ODC)
In a context where transparency and data protection frameworks are still developing, ODC starts from building trust with government partners, and strengthening capacity for data management, integration, and sharing.
🇫🇷 France | CroissantLLM
Built on cross-ministry datasets from France’s government open data platform, CroissantLLM trained a truly bilingual French–English open language model with 1.3 billion parameters, combining transparency with strong performance.
Full report: see first comment ⬇️