MUNI AI Guidelines: Reference Document for E1 and E2 Mini-Books

Research compiled March 2026. All URLs verified at time of research.

The short version for workshop use

MUNI has non-binding recommendations (not binding regulations) on AI use. The core rule is simple:

Undisclosed AI use that affects content = treated as plagiarism (equivalent to ghostwriting).

Everything else follows from this: if AI touched the content of your work, you must say so, how, and with what tool.

There is no MUNI policy for researchers — only for students and teaching. Researchers follow journal and funder policies independently.

Primary MUNI documents

What is permitted, required, and prohibited

Permitted without disclosure

• Grammar, style, and spelling checkers (Grammarly, LanguageTool, etc.)

• Machine translation (DeepL, Google Translate) — note the tool in citation

• Citation management tools

• Academic database search (Google Scholar AI features, Elicit, Consensus, Scopus AI)

• AI-powered summarisation of long texts for personal reading/orientation

• Structure suggestions and content outlines

• Locating published sources

Permitted with mandatory disclosure

• Generative AI (Claude, ChatGPT, Copilot, Gemini) where output materially affects the content of written work — essays, term papers, theses

• AI-assisted argumentation, conclusion drafting, or text formulation

• Any case where AI contributed to what is submitted as your own intellectual work

Restricted or prohibited

• Undisclosed AI use where content is affected (= plagiarism / ghostwriting)

• Using AI as a substitute for reading and engaging with original sources

• Presenting AI-generated content as independently authored

• Inputting personal data, confidential documents, or sensitive research data into non-approved tools

The four accepted disclosure methods

MUNI's recommendations accept four ways of disclosing AI use:

1. General declaration — a statement at the beginning or end of the work that AI tools were used, specifying which and for what purpose
2. Direct in-text citation — citing the AI tool inline, like any other source
3. Methodology section — describing AI use in a dedicated methods or transparency section (common in academic papers)
4. Appendix — attaching the full prompt history and AI outputs as an appendix

Recommended thesis declaration (MUNI's suggested wording):

"I declare that I have used AI tools in accordance with the principles of academic integrity and that I have made appropriate reference to the use of these tools in the thesis."

Recommended in-text citation format:

"In writing this part of the thesis, I have used the tool [name and version, URL] for the purpose of [description]. The tool was prompted with [text of prompt] on [date]."

MUNI-approved AI tools (IT portal)

Claude, ChatGPT, and other external tools: Not in the approved list, not prohibited for non-sensitive content. For sensitive data (personal data, confidential research material, unpublished work), use e-Infra or other institutional platforms only.

MUNI explicitly discourages commercial AI-detection tools (TurnItIn AI detection, GPTZero, etc.) due to accuracy concerns and potential student privacy violations.

What is NOT covered (gaps as of March 2026)

These gaps are important to communicate in both mini-books:

No researcher/publication policy MUNI has no guidance on AI use in writing research articles, grant applications, data analysis, or peer review. Researchers operate according to their journals' and funders' own policies. When in doubt, disclose.

No PhD-specific guidance PhD students are in a grey zone — they are both students (subject to student rules) and researchers (no specific guidance). The safe default: apply student disclosure rules to any submitted work; follow journal/funder rules for research outputs.

No binding regulations The Study and Examination Regulations (SaZŘ, September 2024 version) contain no AI-specific provisions. AI policy is non-binding recommendations only. Violations are handled under general plagiarism and academic integrity rules, which can create ambiguity.

Instructor decides per course MUNI's approach is deliberately decentralized: "The possibility or degree of use of AI always depends on the specific nature of the assignment and the instructions of the lecturer." This means rules vary significantly across courses, faculties, and programmes. Students must ask their instructor.

The awareness problem

Research involving MUNI academics found:

• 49% of Czech university students were unsure whether their university had any official AI rules

• 15% were aware guidelines existed

• 23% knew rules existed but did not understand their content

This means a significant proportion of MUNI students are currently operating without clear understanding of the rules — even rules that have been in place since 2023. The mini-books can help address this directly.

Implications for the mini-books

E1 (scholars)

• No institutional AI policy covers research output → researchers must self-govern by journal and funder policies

• Data security rule is binding: never input sensitive research data (interview transcripts, personal data, unpublished findings) into external AI tools. Use e-Infra or Claude Code (local processing, data never leaves the machine)

• FF MU library has two AI methodologists available for consultation — a resource worth mentioning

• The ethical framing: MUNI emphasises transparency and personal responsibility, not prohibition

E2 (students — all levels)

BC:

• Disclosure rule is the core thing to know: if AI affected your content, say so

• Four disclosure methods — choose the one your instructor prefers

• Ask your instructor before using AI in any graded work

• Machine translation and grammar checking: no disclosure needed

MA:

• For thesis work: use the recommended declaration wording

• In-text citation format is available and recommended for specific AI-assisted sections

• e-Infra platform for any work involving sensitive data (primary source material, interview data)

• No approved AI tool replaces reading original sources

PhD:

• Grey zone between student and researcher rules

• Safe default for submitted academic work: student disclosure rules

• Safe default for research outputs: follow your target journal's AI policy

• For data analysis involving personal or sensitive data: institutional platforms (e-Infra) or Claude Code (local) only

• MUNI has no policy governing AI in grant applications — check GAČR and ERC guidance directly

CUNI researcher recommendations (co-authored by MUNI)

Since MUNI has no standalone researcher AI policy, the most applicable guidance is the inter-university document co-authored by MUNI:

→ See E.cuni-researcher-recommendations for full text

Key addition over MUNI's own documents:

• Qualitative research specific: publish prompts and coding schemes

• Model parameters: document hyperparameters that may influence results

• FAIR principles apply to AI-assisted research

• Explicit warning about AI widening inequalities between well-resourced and less-resourced institutions

Related

• E.cuni-researcher-recommendations — full CUNI PDF, freely distributable, co-authored by MUNI

• E.mini-book-scholars — E1 plan: where these guidelines fit

• E.mini-book-students — E2 plan: how to present rules at each level

• A.critical.limitations — hallucination, data integrity, and the responsible use framing

• A13.examples-dissinet-usecases — data sensitivity note specific to DISSINET

• A.setup.settings-local — Claude Code runs locally; relevant for sensitive data handling