What happened
TNO developed GPT-NL, a sovereign, open-source language model for the Dutch language. Funded by a €13.5 million investment from the Netherlands Enterprise Agency, the model is trained from scratch using strict data criteria to ensure trustworthiness, transparency, and control over data provenance. Model weights are available under a controlled licence, and the project includes a revenue-sharing model for data providers.
Why it matters
Digital sovereignty for the Netherlands and Europe strengthens control over critical AI infrastructure, reducing reliance on non-European providers. For procurement teams, this creates a domestic alternative for language model acquisition, potentially shifting unit economics by offering transparent licensing and data governance. Security architects gain assurance from the model's from-scratch training and strict data exclusion policies, mitigating risks associated with unclear data provenance and intellectual property. This follows the Netherlands' earlier testing of its national AI model, reinforcing a strategic commitment to independent AI development.
