Journal and University AI Policies: How to Read Them When Every Rule Differs

A lecturer submits two papers to journals from two different publishers. The first requires disclosure of any AI tool in the methods section. The second says basic language assistance need not be declared but bans AI-generated figures. The course he teaches follows a department policy, and the thesis students he supervises have their own guidance. There is no single AI rulebook; there are overlapping layers of rules, and the narrowest layer usually decides.

This article does not list individual policies, because they change constantly. Instead it sets out the stable common ground, how to find the rules that actually apply to you, and what to do when they are unclear.

Where almost all publishers agree

• AI is not an author. COPE's position statement and the policies of many major publishers say AI tools cannot meet authorship criteria because they cannot take responsibility for the work.
• Authors are fully responsible for all content, including AI-assisted parts: accuracy, citations, plagiarism.
• Substantive AI use must be disclosed, typically in the methods, acknowledgements or a dedicated statement.
• Reviewers must not upload manuscripts to external AI tools, because manuscripts are confidential.

Where policies differ

These differences are exactly why you need the policy of the journal receiving your paper, not the one you remember from your last submission.

Universities: three layers

1. Institution-wide academic integrity rules, usually stating the principle that unauthorised AI use in assessment is misconduct.
2. Faculty or programme guidance, more specific and sometimes with separate thesis rules.
3. Course or assignment rules set by the instructor, which are usually the deciding layer.

Many universities classify assignments by permitted use: no AI; AI for specific tasks such as brainstorming or language checks; open use with disclosure; or tasks designed around using and critiquing AI. If an assignment brief does not say which applies, that is the moment to ask.

Funders and grant applications

Several funders now issue guidance for both applicants and assessors. The recurring points: applicants are responsible for the accuracy and originality of their proposals, and assessors must not put applications into AI tools because they contain confidential, unpublished ideas. If you both apply and review, the two sets of rules govern two different roles.

Finding the rules that apply to you

• Journals: the author guidelines, usually in a section on AI or within publication ethics. If the journal points to a publisher-wide policy, read both.
• Courses: the syllabus, the assignment brief and the course page in the learning management system.
• Theses: graduate school regulations, thesis formatting guidance and any agreement with your supervisor.
• Funders: applicant guidance and assessor guidance.

Note the date you read a policy and save a copy. AI policies are updated more often than almost any other rules, and the version you read when you started writing may differ from the one in force when you submit.

When rules are vague or conflict

1. Ask in writing, describing exactly what you plan to do: “I intend to use an AI assistant to check the grammar of English text I have written myself. Is this allowed, and does it need to be disclosed?” Specific questions get specific answers.
2. Keep the reply with your project records.
3. If two layers conflict, follow the stricter one, or ask the more specific layer (the course instructor, the journal editor) to confirm.
4. If you cannot ask, choose the most transparent option: disclose fully. Over-disclosure almost never hurts.

Common misconceptions

• “If it isn't banned, it's allowed.” Many integrity policies predate generative AI yet still apply: submitting work that is not your own is misconduct, whoever or whatever wrote it.
• “The journal didn't ask, so I don't need to say.” Norms are shifting, and voluntary disclosure protects you if rules tighten later.
• “Features built into my word processor aren't AI.” Rewrite and auto-draft features are generative AI. Judge a tool by what it does, not what it is called.

A five-minute checklist

1. Have I read the current AI policy of the exact venue receiving my work?
2. Do I know whether what I did with AI is permitted, requires disclosure or is prohibited?
3. Did I use AI to create any images, and does the venue allow that?
4. Have I disclosed in the right place and at the right level of detail?
5. Do all co-authors know about and agree with the disclosure?

Your next step: open the author guidelines of the journal you plan to submit to, find the AI section and paste it, with today's date, at the top of your project folder. When you write your disclosure, you will have the exact basis in hand.

Câu hỏi thường gặp