European "institutional" projects are always so brain fucked...

If you look "how to apply they say": We've kept the application form short so as not to waste your time.

They you see the form... https://nlnet.nl/propose/

ok, getting thousands of euros deserve a little bit of paperwork, but let's not call that form "short".

But what triggered me is the following field of the form: Which model did you use? What did you use it for? Please submit the dates of the prompts, the prompts themselves and the unedited output in this text field.

Then you can see the "AI" policy that is attached: https://nlnet.nl/foundation/policies/generativeAI/

And in my modest opinion, that is another perfect example of the bureaucratic stupidity nightmare.

I would understand to ask details about your intended usage of LLMs for the project. Also to ensure that LLMs is only used as an assisting tool but not to do the work autonomously.

But imagine listing the "dates" and "prompts" that you used, for example to ask the llm to correct your badly written sentences to fill the stupid fields of their form?

And then, recording every time you used a LLM (coding or not) for whatever for the project... Assuming that the recipient of the grant will not lie anyway. This record doesn't give any proof that the code is not "stolen", "copied", or "contaminated".

Just have the delivery evaluated. Independently or by the project receiving the change. Judge on results for once instead of getting orgasms based on the paperwork trail. In the end, why should we care if the author had to use a LLM, a code editor, a spreadsheet, a marabout, or whatever else? If we can confirm that the result is there, in satisfactory condition (clean and safe code, good performance, ...) and without apparent "copyright violation".