Initiative overview
SPARK is a central project of the Federal Government to modernise public administration through artificial intelligence. It addresses challenges in planning and approval procedures. The software was developed through an innovation partnership with the Federal Ministry for Digital and State Modernization developing the solution and the Federal Ministry for Economic Affairs and Energy initiating and financing it as the department responsible for the development of Germany's hydrogen infrastructure.
The modules are modular, scalable and transferable to a variety of administrative procedures. First agents are already working in Hamburg for the approval of hydrogen core grid lines. The modules published so far cover the extraction of relevant information from application documents, formal completeness and plausibility checks, and supporting functions for process overviews, subsequent submissions and comments. Their basis is AI-supported legal dogmatics connected to legal databases, enabling automated deconstruction of norms and legal evaluation; further modules on substantive examination, legal assessment and decision-making are to follow during the year.
Following the "Best Use of AI in Government Services" award at the World Government Summit in Dubai in February 2026, the modules were made available under the "Public Money, Public Code" principle, allowing municipalities, companies and civil society to use and further develop them without licence fees. A Parliamentary State Secretary stated that approval durations could be halved, while the Federal Digital Minister referred to an estimated 80 percent acceleration of approval procedures based on the project's stage in 2026. For citizens and companies, this is stated to mean shorter waiting times, more transparency and more reliable processing.
Human oversight is central: the AI prepares information and proposals, while the final decision and legal responsibility remain with the responsible employees. The published code is to be understood as a reference and basis for integration rather than a finished security configuration, and a dedicated security review is stated to be mandatory before the AI is involved in productive approval procedures.
