Ethical Use of AI in Academic Research

This draft proposes a code for the ethical use of AI in academic research that rests on the obligations of the author, not on detecting AI. It is published to open a discussion, and comments, objections and amendments are welcome. Current guidance agrees on two points. COPE holds that AI tools cannot be listed as authors because they cannot take responsibility for the work. The ICMJE recommendations of January 2026 require authors to disclose the tool, its purpose, and their own verification of the content. These rules say what to declare, but not what the author's responsibility consists of in practice, and enforcement still leans on AI detectors whose reliability remains contested. This code takes a different route. It treats AI as a tool, much like a calculator: a force multiplier when the researcher frames the problem, and a source of shallow or false output when the tool is left to frame it. The code does not ask whether AI was used. It asks whether the author met the obligations that authorship already implies, and it spells those obligations out so they can be checked. The code defines only what is prohibited. What is permitted, and any exceptions, are left to journals, institutions and their ethics committees. These guidelines are not binding regulation. They build on the ICMJE recommendations of January 2026, turning them into concrete obligations, and are addressed to institutions and journals seeking an informed AI policy. The assessment criteria are meant for practical use: OpenPlag, the plagiarism detection platform developed by the author, is being developed to implement them.