Artificial Intelligence (AI) in Research

Researchers remain fully accountable for any content generated using AI tools; such tools do not diminish their responsibility for the integrity of their research (European Commission Living Guidelines on the Responsible Use of Generative AI in Research, 2026).

• Before using AI tools, researchers should assess their risks, for example, regarding handling data, and ensure their suitability and institutional approval status.
• AI tools must support, not replace, independent scientific judgment.
• Researchers must always maintain their meaningful oversight when using AI tools. They must critically assess and validate AI-generated outputs before using them in research.
• AI tools can’t be listed as authors or co-authors. Authorship implies accountability, which cannot be delegated to an AI tool.

AI-generated outputs may be inaccurate, fabricated, or misleading. Researchers are therefore responsible for independently verifying the reliability of any AI-generated content before use.

• AI-generated outputs must not be treated as reliable sources of information, and where feasible, should be checked against independent sources.
• Researchers must always critically evaluate the AI-generated outputs for correctness, accuracy, and relevance.
• Any limitations or uncertainty introduced through AI use must be acknowledged.

Researchers must protect personal, sensitive, and confidential information when using AI tools. Entering such data into AI tools can raise privacy or security concerns, as some tools may use it to train and improve their models.

• Researchers must not enter any personal data, as defined by the GDPR, into AI tools unless they are certain that the tool is GDPR-compliant and that a valid legal basis exists (e.g., the data subject's explicit consent).  
• Researchers must also avoid sharing confidential and sensitive data (e.g., security-sensitive or export-controlled research materials) with AI tools.
• When processing such information with AI tools is necessary, researchers should use only institutionally approved tools and ensure that applicable data protection, confidentiality and security requirements are met.
• Where possible, researchers should opt out of their data and the outputs from being used for improving the AI model. To do so, they should check the AI tools’ policies to learn the default settings and how to change them.
• Any data breach or unintended sharing of personal or confidential information must be reported to CERT-UT at cert@utwente.nl. 

Researchers must ensure that both the input to AI tools and the outputs generated comply with copyright and licensing requirements.

• Researchers are responsible for the originality and legality of AI-generated outputs they use or share. AI outputs may reproduce existing works, and researchers must check for potential plagiarism or copyright infringement (e.g., in text, code, images). Where sources can be identified, they must be properly cited.
• Researchers should be aware that not all AI tools can generate references to the sources used. Additionally, the references generated by AI tools should be verified for correct citation.
• Researchers must check the license of the materials before uploading them to AI tools. Copyrighted material must not be shared with AI tools unless the license or other legal conditions permit it. The purpose of use (e.g., summarisation or translation) does not make such use allowed.
• Researchers should be cautious when uploading their unpublished work, such as manuscripts, research proposals, or any novel ideas to AI tools, as they may lose control over the materials if the AI provider uses them for model improvement. When researchers share copyright with others (e.g., in co-authored works, externally funded projects), they must ensure that they have the necessary permission before sharing the material.
• AI-generated content with minimal human input is generally not protected by copyright. Copyright may apply to the researcher’s own creative contributions.

Further support via Copyright Advisor cip@utwente.nl

The use of AI in research may raise ethical concerns, including bias and societal and environmental impacts.

• AI tools may reflect biases in their training data and/or in how they interpret it. Researchers should be alert to the potential bias or discrimination in AI outputs and identify, mitigate, or report them when they affect research outputs.
• If research participant data will be processed by AI tools, researchers must seek consent and provide the right to withdraw.
• Researchers should consider the broader environmental impact of AI use (e.g., energy consumption) where relevant.
• Researchers should consider whether the use of AI raises ethical concerns, for example, related to human participants, bias, environmental impacts, or dual-use aspects. Where these concerns are identified, researchers should consult the relevant ethics committee and submit an ethics application if needed.

Researchers must transparently disclose substantial use of AI in the research process (European Commission Living guidelines on the responsible use of generative AI in research, 2026).

• Researchers should document details of substantial use of AI, including the tool and version, configuration, parameters, and interaction context (e.g., prompts). This documentation supports reproducibility and accountability.
• The role of AIs should be clearly described in research documentation and publications.
• Any limitations that the use of AI poses on the reproducibility must be acknowledged.
• Disclosure of AI use must be in accordance with disciplinary standards and publisher or journal requirements. Some journals require minimal information for AI disclosure, while others may request comprehensive documentation of AI use, including prompts and outputs, as supplemental materials.
• The level of disclosure should be proportionate to the extent of AI use (e.g., minor text editing vs. substantive text generation).
• Disclosure should include, where relevant, the name of the AI tool, the version, the purpose for which it was used (e.g., data analysis, image generation, text drafting, etc.), and the extent of AI assistance.
• The disclosure can be included in the method section, the acknowledgment, or a dedicated section for AI disclosure, depending on the extent of use and publication requirements.

Researchers should make informed and responsible choices when selecting AI tools for research. They should select tools that are suitable for the intended research task, while considering the following:

• Use AI tools that comply with applicable regulations such as the EU AI Act and GDPR. UT is developing an AI Compliance Framework to provide guidance on legal and regulatory matters: AI and IT Security
• Choose AI tools that ensure strong data protection and allow control over data storage, reuse, and history. For example, use AI tools that provide options to disable data retention or the use of input data for model training.
• Select AI tools from trustworthy and transparent providers. Consider the provider’s reputation, accountability, and policies regarding data use, security, and model development.
• Select AI tools that provide sufficient transparency about, for example, the sources, data, and methods.
• Prioritize institutionally approved, licensed, or locally hosted AI tools.
• Consider the environmental impact of the AI tool and, where appropriate, select computationally efficient tools that meet the intended research needs.
• The selection shouldn't be affected by commercial considerations or solely tools availability.

Possible uses of AI

• Brainstorming research questions
• Exploring potential methodologies

Responsible use

• Critically evaluate outputs and check feasibility.
• Ensure originality and take responsibility for the research.

High-risk or prohibited practices

• Delegating the formulation of research questions or hypotheses to AI without intellectual contribution.
• Using ideas without checking their originality and without acknowledging sources.

Possible uses of AI

• Finding relevant publications
• Summarising papers

Responsible use

• Consult the original sources.
• Verify references.
• Scan broader sources to avoid potential bias in AI-based literature review.

High-risk or prohibited practices

• Citing unverified or non-existent sources.
• Copying AI-generated summaries without consulting original sources.

Possible uses of AI

• Transcribing interviews and recordings
• Generating synthetic data
• Annotation or classifying data

Responsible use

• Clearly distinguish between data that were observed, collected, or measured and data that were generated by AI tools.
• Assess and mitigate potential bias in AI-generated data.
• Consider potential harm to individuals or groups.
• Be transparent with research participants about the use of AI and obtain informed consent where required
• Disclose AI use.

High-risk or prohibited practices

• Presenting AI-generated data as real.
• Using AI-generated data without assessing bias.
• Using AI-generated data that harms or misrepresents people or groups.

Possible uses of AI

• Data cleaning
• Writing scripts for analysis
• Data visualizations

Responsible use

• Understand methods and validate outputs.
• Disclose AI use when it influences outputs or interpretation.

High-risk or prohibited practices

• Uploading personal, sensitive, or confidential data without authorisation.
• Generating research data without disclosure.

Possible uses of AI

• Generating and debugging code
• Translating code to other languages

Responsible use

• Review the code, test, and verify its functionality.
• Disclose AI use when it contributes substantially to code generation.

High-risk or prohibited practices

• Using AI-generated code in research without acknowledging AI assistance when substantial.

Possible uses of AI

• Language editing and grammar correction
• Translation
• Improving clarity of the text
• Drafting text

Responsible use

• Critically review and revise the content for accuracy and meaning.
• Disclose AI use when it contributes to drafting or restructuring substantive parts of the text.

High-risk or prohibited practices

• Claiming AI-generated text as if it were human-written.
• Listing AI as an author.

Possible uses of AI

• Generating illustrative figures and tables

Responsible use

• Check if outputs accurately represent the underlying data.

High-risk or prohibited practices

• Generating synthetic figures and images without disclosure.
• Creating figures that misrepresent data or results.

Possible uses of AI

• Language editing of human-written text and improving clarity and structure, if it is permitted by the funder.

Responsible use

• Follow funder requirements regarding AI use.
• Critically review AI-assisted content.
• Protect unpublished proposals, research ideas, and confidential information.

High-risk or prohibited practices

• Using AI to generate substantive parts of a proposal, such as core arguments, research aims or hypotheses, and methodology.
• Uploading unpublished proposals, research ideas, or personal and confidential information to AI tools.

Possible uses of AI

• Improving clarity of review comments, if it is permitted by the journal or funder.

Responsible use

• Follow journal or funder requirements regarding AI use.
• Maintain confidentiality of manuscripts and proposals.
• Protect unpublished proposals, research ideas, and confidential information.

High-risk or prohibited practices

• Uploading unpublished manuscripts or proposals to AI tools. 
• Using AI tools for the assessment of manuscripts or proposals.