The study is a collaborative report published by the Centre for Digital Governance and Sopra Steria and co-authored by Professor Gerhard Hammerschmid.
Open-source AI has the potential to transform public administration – but is that good news? A new study indicates that using open-source AI presents both opportunities and dangers for public administration. The study was co-authored by Professor of Public and Financial Management Gerhard Hammerschmid, alongside Nicholas Robinson, Magdalena Steinhauser and Bernd Peper, and is the result of a collaboration between the Hertie School’s Centre for Digital Governance and consulting firm Sopra Steria.
Based on 32 interviews with public sector leaders in Germany, Australia and Canada, the study identifies growing awareness around open-source AI as well as significant barriers to its adoption. While proprietary AI models dominate today’s market, governments relying solely on them face challenges including vendor lock-in, reduced sovereignty, and rising costs. At the same time, open-source AI comes with its own hurdles, such as a lack of vendor accountability and internal capacity gaps.
The study outlines actionable recommendations to strengthen the openness and resilience of public AI systems, from promoting inter-agency collaboration to increasing procurement flexibility and transparency.
The results of the study will be presented and discussed in an event with experts from government and academia on 17 September 2025 at the Merantix AI Campus Berlin, followed by a panel discussion and networking exchange.