Policy on the use of Generative Artificial Intelligence for NLnet-funded projects
Version 2.0 (October 20, 2026) — version history
Introduction
Generative Artificial Intelligence (GenAI) is rapidly emerging across the ecosystem. Wherever used, it is essential that it be used responsibly and transparently.
This document describes NLnet Foundation's policy on the use of GenAI — including large language models, diffusion models and code assistants — in applications submitted to NLnet and in all work delivered within NLnet-funded grants. It covers both the application process and the execution of funded activities. The policy exclusively concerns GenAI and does not regulate other forms of automation.
Failure to comply may result in rejection of the proposal, submitted work not being accepted for payment, or ultimately in the termination of the running grant.
GenAI within the application process
It is not allowed to use generative AI during the writing of proposals, nor when providing answers during the interactive evaluation of proposals. This includes LLM generation of ideas, as well as restructuring, shortening and translation. Trust in your knowledge and skills, and use your own words. We don't care about spelling or grammar, only about your ideas.
Accessibility exception: there is one noted exception where using AI in the application process is allowed, which is in case it is needed for accessibility purposes. People with e.g. a visual or motor impairment who rely on AI-based assistive technology are allowed to use AI. Like any other use of AI, it must always be made transparent. Instructions on how to disclose use of AI when submitting proposals are provided in the proposal form.
In your application you must disclose if your proposal builds on an existing codebase that already includes a significant amount of code generated by AI, or expects to gain a significant amount of such code during the lifetime in other ways than through the grant. Please also note the recommendations found below on transparency.
GenAI within your project
It is not allowed to use generative AI for producing artefacts of any kind, including code, design, and documentation.
It is allowed to employ GenAI in a read-only manner — meaning, for instance, that you can have an AI tool check your code and identify security issues and flaws, but that you craft fixes and mitigations yourself, and write your own code. Even if you use genAI in a read-only manner, you should always be transparent about it.
Technology does not replace human responsibility for project quality, technical choices and thoroughness. Human contributors remain responsible for correctness, clarity, and reproducibility, and are expected to understand and be able to explain design and code decisions.
Costs incurred for AI usage are not eligible for reimbursement.
Transparency
Provide at the very minimum a high-level description of all known (past and present) use of GenAI. This means summarising the general stance towards generative AI at the project level in the README, man/info pages, and your website. Even better in terms of transparency is publishing your own policy for contributors, outlining do's and don'ts with regards to the use of GenAI — and of course upholding that policy.
As a matter of project hygiene, we additionally recommend (so this is not mandatory but a 'best practice') a per commit disclosure:
- Ask all contributors to disclose at the granularity of individual contributions which model(s) was/were used, which version(s), and a brief description of how GenAI assisted the contributor.
- Make this information available in machine-readable format in the commit logs.
- Make sure provenance data is stored in a logical place where it can easily be found, and avoid hosting it on third-party platforms that require a log-in or may disappear over time.
- Consider choosing a development tool that helps you create commits, and auto-fills the relevant information.
- Consider that the use of proprietary/as-a-service AI may create a social lock-in for other (current and future) contributors to your project, and may create a long term dependency.
Scope of this policy
This policy explicitly deals with Generative AI only (such as Large Language Models and diffusion models). NLnet is a strong proponent of automation and of deterministic and reproducible generation of source code, formal and symbolic proofs, etc. based on strict specifications and scientific and engineering rigour. Similarly, NLnet does not in any way seek to prevent the use of other forms of machine learning, fuzz testing or other beneficial use cases.
Version control
| Date | Version | Comment |
|---|---|---|
| December 8, 2025 | Version 1.0 | Initial release |
| January 26, 2026 | Version 1.1 | Adjustments based on feedback from grantees |
| October 20, 2026 | Version 2.0 | Restrict to read-only |
