Guidelines for Safe and Responsible Use of Artificial Intelligence for Applicants

The guidelines are based on the principles set out in The European Code of Conduct for Research Integrity (PDF, 2023, p. 5) and in ERA Living Guidelines on the Responsible Use of Generative AI in Research (PDF – Fourth version, October 2026).

We also encourage applicants to follow guidelines on the responsible use of AI in research issued by their own institution and by the National Research Ethics Committees. 

Guidelines

1. Reliability in ensuring the quality of research, reflected in the design, methodology, analysis and use of resources. 
   
   This includes aspects related to verifying and reproducing the information produced by the AI for research. It also involves being aware of possible equality and non-discrimination issues in relation to bias and inaccuracies.
   
   
2. Honesty in developing, carrying out, reviewing, reporting and communicating on research transparently, fairly, thoroughly and impartially.
   
   This principle includes disclosing that AI has been used.
   
   
3. Respect for colleagues, research participants, research subjects, society, ecosystems, cultural heritage and the environment.
   
   Responsible use of generative AI should consider the limitations of the technology, its environmental impact and its societal effects (bias, diversity, non-discrimination, fairness and prevention of harm).
   
   This includes proper management of information, respect for privacy, confidentiality and intellectual property rights, and proper source referencing.
   
   
4. Accountability for the research from idea to publication, for its management and organisation, for training, supervision and mentoring, and for its wider societal impacts.
   
   This includes establishing clear responsibilities for AI use within the research team, and maintaining human oversight of decisions and outputs. 

Examples of good practice

To help our applicants use AI in accordance with these principles, we provide here with some examples of good practices. 

Because AI technologies—and its use in research—is continually evolving, the examples below should not be regarded as exhaustive guidance. If in doubt, please refer to the principles above. The actions needed to comply with these principles will depend on the AI tools you use.

We aim to update the examples below as the technology and standards for responsible use in research evolve. 

1. Reliability 

Inaccuracies in generative AI outputs can arise from two main sources: hallucinations and biases in the training data or system instructions. Here, “hallucination” means the generation of false, nonsensical, or inaccurate information by large language models (LLMs) or other generative AI systems. 

Good practice: AI-generated references may be inaccurate or fabricated. Check factual claims against primary sources that you have identified yourself. 

Most commercial LLMs and generative AI-systems has built-in filters, designed to stop it from providing output that is offensive, immoral, or otherwise unacceptable. In addition, these tools rely on pattern-finding mechanisms where they use what they’ve seen before (training data) to draw conclusions about new input. 

For example, when asking an AI service to describe a doctor’s main tasks, the AI may assume that the doctor is female if its training data predominantly contains information about female doctors. Important information about male doctors may then be overlooked. Similarly, English-language research may be over-represented in a literature review because English-language sources dominate the training data, and highly cited papers may be favoured.

In brief: Be critical when assessing the quality of any output: is it complete and unbiased?

2. Honesty 

Research integrity is one of the cornerstones of good research practice includes being honest about any use of AI. The Research Council of Norway is currently developing a mechanism for applicants to declare their use of AI in preparing their project proposal. 

3.Respect 

The use of AI tools may have unintended consequences that a user should be aware of. Information from a proposal, be that novel research ideas or information about planned collaborations, may suddenly be out in the public domain if the text is uploaded to an AI without appropriate safeguards. 

Good practice: Only use AI tools you can trust: either because they’re provided through a trustworthy institution such as a university or research institute, or because you have entered a license agreement with them that allows you control over your information.

Another unintended consequence of the use of AI tools may be their impact on the environment and society. Most AI tools are based on data centres, which consume vast amounts of electricity, water and space. These effects depend on the specific tool used, and how the provider has implemented the service. Before using AI-tools, researchers should consider these risks and effects.

In brief: Consider the effects that using AI tools has outside of your proposal.

4. Accountability 

Above all, a user is responsible for their use of AI and its results. An AI should never be left in charge of any part of the work involved in drafting a proposal. Even seemingly trivial tasks may carry risks: without monitoring the AI and checking its output, you may even run risk of plagiarism – which you as the proposal owner would be responsible for.

Good practice: Make sure you understand what the AI is doing at each step, and that you can explain its actions and effects on your work. Check the original sources to be sure you are not plagiarizing someone else’s work.