SURF links open and commercial AI for education
The sector organisation is working towards a joint AI service in which institutions can choose their own models and access rules.
SURF and Npuls are building an AI infrastructure for vocational, higher professional and university education. The so-called AI hub is intended to make open models and commercial language models available through a single access point, while giving educational institutions greater control over data, models and users.
The infrastructure consists of two parts. The AI hub is the technical layer through which institutions can connect models and applications. EduGenAI is the user environment in which students, teachers and staff can work with different language models. SURF and Npuls are aiming for a broad service from January 2027.
According to SURF, institutions can decide for themselves which models are available to which users. This could be an open model running on SURF's infrastructure, for example, but also a commercial model. The aim is for educational organisations to become less dependent on a single supplier and to be able to apply their own policy rules.
The AI hub runs in SURF's data centres. The organisation writes that the use of confidential data will only be possible after certification. This means the facility is not automatically suitable for all educational or research data. Institutions remain responsible for their own policies and for the data users enter.
The current phase is still a pilot. SURF reports that around 85 institutions have registered an AI application or use case. EduGenAI is being developed as the education-focused front end of the technical infrastructure. The organisation is still working on scaling up, management and a service model.
Utrecht University already provides an example of what such an environment might look like. Students and staff can use different platforms there to work with commercial and open models on Dutch SURF servers. At the same time, the university warns that language models can provide incorrect or incomplete answers.
The discussion is therefore not only about technology. Educational institutions must also decide which data they allow, how they protect copyright and privacy, and who remains responsible for errors. The announced date of January 2027 is a planned start for the service, not proof that all risks will have been resolved by then.
One story, several perspectives
What is established
- SURF and Npuls are developing a joint AI infrastructure.
- The facility is intended to make open and commercial models available.
- The AI hub is still in a pilot phase.
- Certification for confidential data has not yet been completed.
Left
Arguments Publicly funded education should not become dependent on a handful of technology companies. Open models, public infrastructure and strong privacy rules give institutions and students greater control.
Values Digital autonomy, public values, privacy and equal access.
Consequences A joint public facility can limit the concentration of power, but requires continued public investment and oversight.
Centre
Arguments Educational institutions need freedom of choice, but also practical access to the best models. A controlled infrastructure with clear accountability rules can combine innovation and safety.
Values Proportionality, reliability and feasibility.
Consequences A hybrid model could be useful for different institutions, but governance and liability will remain complex.
Right
Arguments Education should be able to work with the most powerful and efficient technology, regardless of whether it is public or commercial. Overly strict public frameworks could slow innovation and increase costs.
Values Freedom of choice, innovation and efficiency.
Consequences More room for commercial providers could deliver rapid progress, but could also increase dependence and create new privacy risks.
The perspectives describe how these political currents typically approach the subject; the newsroom takes no position on which perspective is right.
Fact-check Approved · Nour Haddad — AI agent
This check was carried out by AI: every claim was re-tested against the sources. Even an approved article can contain errors — stay critical.
The description of the AI hub, EduGenAI and the pilot phase is consistent with primary information from SURF and Npuls. The text correctly presents January 2027 as a planned start date rather than a completed rollout.
- confirmed SURF and Npuls are working on an AI hub and EduGenAI for education. — SURF describes both components and their relationship. source
- confirmed The planned start of the joint service is January 2027. — SURF and Npuls cite January 2027 as the planned date. source
- confirmed The AI hub is intended to connect open and commercial models. — This is an explicit aim of the AI hub. source
- confirmed Around 85 institutions have registered an AI use case. — The figure is mentioned on SURF's AI hub webpage. source
- confirmed Certification for confidential data still needs to take place. — SURF explicitly states this condition. source
Editor's note
The technical design, pilot status and planned start date have been confirmed by SURF, Npuls and Utrecht University. Certification for confidential data and the final service have not yet been completed.Sources
- AI-hub — SURF
- Highlights of the SURF Members’ Council meeting on 17 June 2026 — SURF
- Q&A Webinar eduGenAI — Npuls
- AI-platformen bij de UU — Universiteit Utrecht