GPT-NL: Netherlands builds its own AI language model to cut reliance on US tech giants

The Netherlands is building its own large language model, independent of major American AI providers. TNO writes that the project, called GPT-NL, is being developed by research institute TNO together with SURF and the Netherlands Forensic Institute (NFI). The Dutch government has allocated €13.5 million to the initiative through the Netherlands Enterprise Agency.

The project aims to give the Netherlands and Europe greater control over AI infrastructure. The developers describe this as “digital autonomy”: the ability to decide independently how a language model works, what data it uses, and how it handles sensitive information.

Built from scratch with strict data rules

Unlike many existing models, GPT-NL is trained entirely from scratch. This approach avoids inheriting unclear data origins, copyright problems, or personal data from other models. Before any data enters the training process, the team removes and anonymises personal information, excludes confidential and harmful content, and checks for intellectual property issues.

The source code will be published as open source. Model weights will be available under a controlled licence, allowing the developers to track who uses the model and notify users of changes, for example when a data provider opts out.

A body called the Content Board gives data providers and rights holders a say in the project’s direction. Part of any revenue flows back to content creators, which TNO describes as a fairer model of value sharing.

The team also states it actively works to reduce the energy and water consumption involved in training the model, based on ongoing scientific research.

GPT-NL is presented as proof that high-performing AI and public values such as privacy, transparency, and copyright protection are not mutually exclusive.