FEPS Guidelines
Published in July 2026
The provisions of this Code apply to all publications for which the service provision agreement has been signed after 01/09/2026
Context and definitions
Artificial intelligence, commonly abbreviated as “AI”, is a name covering a wide array of technologies that perform tasks or behave in an “intelligent” manner. Recent developments, especially in the field of so-called “generative AI”, present opportunities, risks and limitations for research and communication contents.
The EU Artificial Intelligence Act defines an “AI system” as “a machine-based system that is designed to operate with varying levels of autonomy and that may exhibit adaptiveness after deployment, and that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments” (article 3).
The word “author” is used in the following paragraphs to designate any person writing research, studies, books, policy briefs, articles and other types of publications for FEPS.
Scope of the guidelines
The present guidelines intend to foster responsible and ethical use of AI tools in the context of FEPS’ research activities and publications. The guidelines should be respected by all authors.
The guidelines primarily apply to the use of generative AI tools for research and writing purposes. They mainly focus on the use of commercially/publicly available AI tools regardless of their scope (broadly applicable or with specific purposes) and type of input/output (e.g. written text, figures, data). They also apply to AI tools that are available in software programs through plug-ins. This includes both online and offline use of commercially/publicly available AI tools (e.g. within a closed environment).
AI tools for data mining and for data analysis that do not require to make data available on an external portal are not targeted in these guidelines. However, their use and methodological implications should still be discussed with FEPS prior to the research process.
The following guidelines are adapted from the guidelines developed by the Vrije Universiteit Brussels and the Living guidelines on the responsible use of generative AI in research developed by the European Research Area Forum. We encourage all authors to read them and to familiarise themselves with current knowledge and debates on the use of AI.
General approach to AI
- As much as possible, FEPS wishes to prioritise contents developed by human beings.
- Authors are responsible for the quality and, when relevant, the scientific standards of their manuscripts. Authors are responsible for verifying the reliability of the AI outputs.
- Authors should openly discuss their intentions to use AI with FEPS at the beginning of the project or writing process.
- While FEPS acknowledges that interactions between authors and AI tools can generate ideas and be useful in some contexts, FEPS equally values the expertise, human experiences, critical insights and creativity of authors.
- FEPS encourages all authors to critically reflect on the balance between the expected benefits of using AI tools and the environmental costs of such use.
Guiding principles
Deploying AI tools in the context of FEPS research and writing activities should always adhere to the following principles.
1. Human first, AI second
The fundamental starting point for FEPS’ view on the use of AI for research and writing is that while AI tools can be legitimate and useful tools at various phases of the research and the writing process, authors should always work with a human first, AI second approach. For FEPS, this means that human beings write the first draft of the text of a publication or other type of deliverable. This can be done based on AI-assisted desk research, brainstorming, etc., as exemplified in Table 1 below. AI can then be used to review work written by humans. Using AI to generate a draft text and then adding a “human touch” by verifying the correctness and coherence of the content or in other ways is not an acceptable practice for FEPS, as this does not constitute genuine human work, even if the source materials on which the AI generated the text were selected by a human.
2. Authors are fully responsible for AI outputs and for the appropriateness of the research process in which AI is used in the case of activities involving research.
Responsibility cannot be passed on to AI tools. In practice, this means that authors can be held accountable for inappropriate use of AI tools, such as failure to comply with the rules regarding scientific and intellectual integrity (plagiarism, citing sources, etc.). This rule is consistent with the position of the Committee of Publication Ethics (COPE) that “authors are fully responsible for the content of their manuscript, even those parts produced by an AI tool, and are thus guilty for any breach of publication ethics”.
3. Authors must provide the greatest possible degree of transparency on the use of AI tools in their research and writing activities.
Authors should discuss the use of AI tools before the beginning of a project with FEPS and must provide an AI use disclosure statement. In the context of research activities, the way AI tools are used often needs to be described in detail in the methodology, Materials & Methods section and/or Supplementary Data, including prompts, output, and name and version of the AI tool.
4. Authors need to check the outputs of AI tools.
For each specific use of AI tools in research and writing activities, authors should strictly respect the “do’s” and “don’ts” identified in Table 1 in the next section. For example, various verification steps must be built in in the research and writing process. One should always look for the original sources and, where applicable, check licenses for the original work. Original ideas must also be properly attributed to their creator. Training data used by AI tools may also be copyrighted or reuse may be restricted by licensing. In that case, one cannot literally copy the AI output. In addition, it is necessary to provide information (mainly facts and figures) to allow fact checking. Researchers should check whether outputs reflect sufficiently diverse views (e.g. conflicting results of studies). In addition, one must check whether scientific statements are sufficiently supported by empirical material (if relevant to the research discipline). The more “autonomous” AI tools work, the greater the degree of verification that authors have to perform. The greater the risk of potential harm and the amount of harm that could be caused, the greater the degree of scrutiny authors must exercise.
5. Authors should carefully consider and mitigate the impacts of the limitations of AI on the quality of their manuscripts.
Authors must be very critical of text outputs from AI tools. Authors must be aware that using AI tools can also be a disadvantage in some circumstances. Using AI tools can involve issues related to biases and inaccuracies. Examples include:
- Training data used by AI tools may exhibit different forms of bias; may be influenced by existing structural inequalities between groups; may contain discriminatory language, racist undertones, and incorrect information and argumentation; may be dated; may contain copyrighted material that has been unlawfully released. AI tools can also generate text with generic statements with limited value.
- It is often impossible to determine exactly how algorithms arrive at a certain output. A lack of insight into the exact workings of AI tools can have serious consequences.
- The outputs of AI tools can be presented as convincing but can be inaccurate, completely wrong or simply irrelevant. The self-confidence with which AI tools make statements can therefore be misleading.
- Generative AI tools create outputs that cannot always be exactly replicated. For example, the same prompt can lead to different text fragments.
- Outputs may contain plagiarised elements.
6. Authors must properly manage information and data used in the context of their FEPS activities, in compliance with relevant European laws.
Research activities carried out for FEPS should respect research participants, privacy and confidentiality. The following data must not be entered in AI tools:
- Personal data: data that can (directly or indirectly) lead to the identification of persons.
- Data that is important with a view to future valorisation or data that is (or can be) protected by intellectual property law.
- Data whose release could be ethically problematic, such as in the event of a real risk of inappropriate use or group harm.
- Data where existing contracts (e.g. non-disclosure agreements) or rules from funding organisations impose restrictions on sharing.
- Data that is protected by copyright or database rights of another person, unless with that person’s permission.
In case of any doubt, the author should contact FEPS and the policy analyst in charge of the project to discuss the matter at hand before the use of AI tools. In all cases, we recommend that researchers check the privacy settings of the AI tools and adjust them if deemed necessary.
7. Authors should not feel obliged to use AI to proofread their work from a linguistic and grammatical perspective
FEPS is aware that many authors are not native English speakers. This is reflected in the distinctiveness and linguistic richness of our written publications. Using AI to correct spelling and grammar is an understandable reflex. However, in light of how AI models work, this ends up homogenising the way English is used and eliminating the linguistic diversity which makes human writing unique. FEPS always works with trusted, human copy editors to check the quality and correctness of written work while preserving the authors’ unique linguistic fingerprint. As such, using AI tools for this is neither recommended nor needed.
Identifying the research purposes for which AI can be used
The guidelines of the Vrije Universiteit Brussels provide a useful classification of the different purposes for which AI can be used while carrying out research and writing activities (see Table 1 in the document attached). Authors working with FEPS should review them, clearly identify their own use of AI and apply identified best practices.
References
European Commission, 2025. Living guidelines on the responsible use of generative AI in research. European Research Area Forum Stakeholder’s Document.
Vrije Universiteit Brussels, 2023. VUB guidelines for responsible use of artificial intelligence for research purposes.
COPE Council, 2024. COPE Position: Authorship and AI tools. Committee on Publication Ethics.