AI ethics & disclosure in research

By Dr. Rafiq Muhammad, MD, PhD · Updated June 2026

Using AI well is partly a craft and partly a set of rules — and the rules are where careers get damaged. Two anchor everything: an AI cannot be an author, and substantive use usually has to be disclosed. Around those sit your journal’s and institution’s specific policies, which you should read before you submit, not after.

Why AI can’t be an author

The ICMJE and COPE positions are clear: AI tools can’t be listed as authors, because authorship requires accountability — approving the final version, vouching for the integrity of the work, and answering queries about it. An AI can do none of those. Human authors stay fully responsible for the entire paper, including any AI-assisted parts. (You can acknowledge a tool; you can’t credit it as an author.)

Disclosure: the safe default

More and more journals and universities require you to declare if and how you used generative AI — usually in the methods or acknowledgements. Even where it isn’t strictly required, disclosing substantive use is the honest default. Policies vary, so check your target venue: some want disclosure only for text generation, others for any use beyond basic spell-check.

How to write a disclosure statement

Keep it specific and factual — name the tool, what you used it for, and confirm you take responsibility:

What counts as misuse

• Passing off AI-generated text or ideas as wholly your own where disclosure is required.
• Citing fabricated references, or fabricating/altering data or results with AI.
• Using AI in ways your ethics approval or journal policy forbids — including pasting confidential or participant data into a tool that retains it.
• Using AI to evade plagiarism or AI detectors — itself a form of academic dishonesty.

None of this is a reason to avoid AI — it’s a reason to use it transparently. The same standards of honesty that govern peer review and citation simply extend to your tools.