AI in grant applications: can nonprofits use ChatGPT? Norway's Research Council rules and five tips
Yes, a nonprofit can use AI to draft a grant application, but the responsibility never moves to the machine. The Research Council of Norway's guidelines for applicants (updated 11 November 2026) require applicants to be open about AI use and state that the applicant bears full responsibility for everything in the application. Here is what that means in practice, plus five rules a board can adopt.
Key facts as of November 2026
- The Research Council of Norway (Forskningsrådet) published guidelines on applicants' use of AI on 2 November 2026 and updated them on 11 November.
- The guidelines rest on four principles: reliability, honesty, respect and accountability.
- The Research Council is working on a self-declaration of AI use for project applications, but has not said when it will arrive.
- The EU's ERA Living Guidelines (fourth version, October 2026) separate substantial AI use, which should be declared, from simple proofreading, which generally need not be.
- We have found no equivalent applicant guidelines from Bufdir, the Norwegian Gaming and Foundation Authority (Lottstift), Stiftelsen Dam or Sparebankstiftelsen DNB as of 15 November 2026.
- 16 per cent of Norwegian voluntary organisations say they have skipped a grant because the documentation requirements were too heavy (Menon Economics, 2026).
Is it allowed to use ChatGPT to write a grant application?
We have found no rules at Bufdir, Lottstift, Stiftelsen Dam or Sparebankstiftelsen DNB that ban AI in the application process. That does not mean anything goes. A funder bases its decision on what the application says, and wrong figures, invented sources or leaked personal data are the organisation's problem – whoever, or whatever, wrote the text.
The Research Council has now put this in writing for its own applicants. It points to a 2025 survey of 3,234 researchers in 113 countries in which 41 per cent were already using AI to prepare applications. The figure covers researchers worldwide, not Norwegian nonprofits.
What do the Research Council's AI guidelines actually say?
The guidelines are the Research Council's description of safe and responsible AI use for everyone applying for funding. Applicants are expected to read them, and they build on European guidelines on research integrity and the EU's ERA Living Guidelines on the Responsible Use of Generative AI in Research.
The four principles
- Reliability: applicants must ensure quality and verify AI-generated information. The guidelines warn that references produced by AI can be wrong or fabricated.
- Honesty: applicants must disclose that AI has been used.
- Respect: privacy, confidentiality, intellectual property and correct citation.
- Accountability: applicants must be able to explain how AI was used, and human oversight of the results is required.
Note the distinction: openness is a requirement, while the advice on primary sources and choice of tools is framed as good practice. The self-declaration is not in place yet either – the Research Council says it is currently working on a solution for declaring AI use in project applications.
Do the rules apply to Bufdir, Lottstift and private foundations?
Not directly. The guidelines cover applications to the Research Council only. Bufdir, Lottstift, municipalities and foundations each have their own rules, and as of 15 November 2026 we have not found AI guidance for applicants from the largest of them.
This is our assessment: the principles work just as well for a small activity grant as for a research project. Following them now prepares you for when other funders ask the same questions.
Why is AI so tempting for grant writing?
Because applications are a real burden for many organisations. In Menon Economics' survey of documentation requirements in the voluntary sector for Lottstift (report 2/2026), 16 per cent said they had held back from applying because the requirements were too extensive. Small organisations struggle most, mainly because of limited administrative capacity.
The picture is mixed. According to the full report, 2,185 of 35,875 organisations responded (a 6 per cent response rate), and around half spent less than five hours on their most recent application. Our guide shows how to find and win more grants.
Five rules for the board when AI helps with an application
The rules below are our translation of the Research Council and EU guidelines into nonprofit practice. Each rule names its source; the practical advice is our own assessment.
1. Disclose the use
The Research Council requires openness. The ERA guidelines add that spellchecking, grammar correction and formatting generally need not be declared, but: “In case of doubt, disclosure is advised.” Our advice: if AI contributed substance – the project description, objectives or budget narrative – add one sentence saying so in the application or cover letter.
2. Check everything against primary sources
Check every figure, quote, legal reference and deadline against its source and the call text. The Research Council recommends checking claims against primary sources you have identified yourself. AI can invent figures and references that sound entirely plausible.
3. Keep confidential information out
The Research Council warns that personal data, new ideas and planned partnerships can become available to third parties if the text is uploaded to an AI tool without adequate safeguards. In our view, names and health details of members and service users, national ID numbers, bank account numbers, internal board papers and partners' unpublished plans should never be pasted into a free tool. Read more in our piece on member data and AI assistants.
4. Keep your prompts
The Research Council asks applicants to be able to account for their AI use, for example by keeping the prompts they used. Save the prompts, tool name, date and user in the application folder.
5. A human reads and signs
The Research Council requires human oversight, and the ERA guidelines state that using AI agents can never be treated as delegating control or responsibility. The board should decide who reads the whole application before submission and who signs it. Then write the rules into your organisation's AI policy.
Don't try tricks with hidden prompts
The ERA guidelines warn funders about applications containing hidden prompts, such as invisible text meant to nudge an AI towards a kinder assessment. Funders may ban it, and it erodes trust.
What is OpenInfo doing in this space?
We have an interest here and want to be upfront about it: OpenInfo develops Grant Finder, an AI agent that finds funding opportunities for nonprofits, and Grant Writer, which drafts applications from an organisation's own material, checks the requirements in the call and includes human quality assurance. Neither tool exempts an organisation from the rules above.
Both will be shown on Friday 20 November, 13:15–14:15, at the free AI day for nonprofits at Microsoft in Oslo, where OpenInfo chair Håkon Berntsen leads the session. The event has 130 seats and is not streamed.
Frequently asked questions
Can nonprofits use AI for grant writing, and do they have to disclose it?
Yes, they can. The Research Council of Norway requires disclosure for its own applications; for other funders, our advice is to disclose whenever AI contributed substance, not just proofreading.
Do we need to disclose AI if we only used it for proofreading?
According to the EU's ERA guidelines, simple spellchecking, grammar and formatting generally need not be declared. If in doubt, or if the funder asks, disclose it.
What should never be pasted into an AI tool?
Personal data about members and service users – especially health details and national ID numbers – as well as bank details, internal board documents and partners' unpublished plans.
Will disclosing AI lower our score?
The ERA guidelines cite research suggesting that declared AI assistance can lead to lower scores, and therefore ask funders to make sure openness is not penalised. That is no reason to hide it.