Protecting Personal Data in Local Government
Protecting sensitive information is a tricky challenge facing local governments today. In the Netherlands, the municipality of Hoeksche Waard has taken an innovative step by developing an anonymisation solution that uses large language models (LLMs) to automatically detect and mask personal data in documents.
“The idea originated from the growing challenge municipalities face with the large volume of information requests under the Wet open overheid (Woo)(Dutch Open Government Act). Much of this information contains sensitive personal data that must be anonymized before publication. Traditional manual methods were too time-consuming and error-prone.” explains Femke Bruinsma, from the Open WebConcept community.
The anonymisation tool, built with open source components and self-learning algorithms, replaces old practices where documents are manually redacted or censored with black bars. The methods currently used by public agencies are costly, time-consuming, and often inaccessible to people with disabilities, such as those using screen readers. By contrast, the LLM-based approach allows municipalities to share information securely while maintaining compliance with the Dutch Open Government Act (WOO) and digital accessibility standards.
De haven van Oud-Beijerland © Martijn van Ruiven
Recognition at the Computable Awards 2025
The project is gaining recognition from the Computable Awards 2025. The Computable Awards, organized annually by Dutch IT magazine Computable since 2006, honor excellence and innovation in ICT across organisations, projects, and individuals, with winners chosen through public voting and expert jury evaluation. The independent expert jury for the Awards gave the initiative a score of 8.25, making it one of the ten highest-rated government projects of the year.
“Our ambition goes beyond solving a local challenge: we want to provide a tool that strengthens the entire public sector. Anonymisation is a shared obligation, and by making the solution open and locally deployable, we combine data sovereignty with broad usability. This nomination underlines that collaboration and open source can deliver real public value and we are proud to contribute to that movement.” - Amanda Hanemaaijer, the project leader on behalf of gemeente Hoeksche Waard
The project has become one of the leading government projects in the Netherlands, standing alongside initiatives by national agencies like DUO and Logius, and fellow municipalities Rotterdam and Breda. Logius, under the Ministry of the Interior (BZK), manages ICT solutions and standards for seamless communication between government, citizens, and businesses. DUO, under the Ministry of Education (OCW), handles student funding, exams, diploma recognition, education financing, and data systems, while also supporting other ministries on integration and childcare.
Government Innovation Through Community Collaboration
At the core, the anonymisation project is a collaboration between Hoeksche Waard, local suppliers, and the wider Open Webconcept (OWC) community. Development is made possible with innovation budgets of the Ministry of the Interior and Kingdom Relations (BZK).
OWC gathers over 40 municipalities and partners in a joint effort to develop reusable digital building blocks. This collaborative structure ensures that tools such as the LLM anonymiser are not isolated, but solutions that can be scaled and shared across the public sector. By releasing the anonymisation technology as open source, other municipalities and organizations can adopt, adapt, and improve the tool without having to start from scratch. This improves continuity, reduces costs, and accelerates innovation across local government.
For civil servants, the benefits are tangible and immediate: less time spent manually reviewing documents and more time available to support residents. For citizens, it means government data is handled faster, more secure, and more accessible.
A broader trend in European public administrations
Similar open-source LLM initiatives are emerging across Europe. In Poland, the Ministry of Digital Affairs is preparing to integrate an AI assistant into the national digital wallet mObywatel by the end of 2025, powered by PLLuM (Polish Large Language Model). Meanwhile, in Switzerland, researchers are developing a multilingual open-source model trained on public infrastructure, designed to support over 1,500 languages.
This article was updated on 1 October 2025, with additional input from the Gemeente Hoeksche Waard and the Open WebConcept community.