Using LLMs Safely

What is this section?

This section provides practical guidance on how to use LLMs in a way that is aligned with institutional rules, data protection requirements, and responsible research practices.

It complements the rest of this guide by answering a different kind of question:

  • Not: how do I use LLMs?
  • But: am I allowed to use this LLM setup for my data and task?

Why this matters

Using LLMs in research is not just a technical decision. It is also a data and responsibility decision.

LLMs are easy to use, but they introduce risks that are not always visible:

  • data may be sent to external providers
  • usage may be logged or monitored
  • institutional policies may restrict certain tools
  • legal frameworks (e.g., GDPR) may apply

As a result, the same task can be: - acceptable in one setup
- problematic in another

What this section helps you do

This section helps you:

  • decide whether your data can be used with LLMs
  • understand what your institution may allow or restrict
  • choose appropriate infrastructure (chat, API, VM, HPC)
  • find relevant policy and guidance documents

Relationship to other sections

  • How to use LLMs in research → explains usage modes, workflows and tools
  • Institutional resources → explains available infrastructure in the Netherlands
  • This section → explains constraints and safe usage

Together, they provide a complete view of: - what you can do
- where you can do it
- what you should or should not do

A simple way to approach LLM use in research

  1. Start with your data

  2. Is it public, internal, or sensitive?

  3. Check institutional rules

  4. Are there restrictions or approved tools?

  5. Choose the right infrastructure

  6. Public tools, institutional tools, or secure environments?

  7. When in doubt, be conservative

  8. Use more controlled environments or ask for guidance