EPFL – École polytechnique fédérale de Lausanne

EPFL guidelines for using AI in studying advise students that they should not use AI during exams or assessments unless explicitly allowed, and tell students to check their course policy or ask their instructor. The guidelines state that assessment material that is not the student's personal and original contribution must be recognizable as such under LEX1.3.3 Article4; they advise disclosing AI-tool use in a statement and checking the course policy or instructor for its formulation.

The French LEX1.3.3 directive, originally effective1 January2013 and revised1 September2026 as version2.2, requires scientific and academic work to distinguish borrowed material from the author's personal and original contribution under Article4. Article4 says that failing to identify borrowed material presents it as a personal and original contribution and constitutes plagiarism, referring to Article8. The official Polylex directory states that only the French version is legally binding and that the English version is for information and has no legal force.

EPFL Library guidance for manuscripts intended for publication states that the LLM used, what it was used for, and how it was used must be disclosed in the manuscript, while adhering to the policies on AI-generated content of the specific journal or publisher chosen for submission.

According to EPFL Library's description of its agreement with Elsevier, open access publications with a CC-BY license may be used with any available AI tool, including for tool development and training, with free sharing of results. For licensed closed publications or Open Access publications under a restrictive CC-BY-NC-ND license where Elsevier owns all or key rights, use is permitted with closed versions of AI tools that do not train the algorithm or learn from/incorporate inputs (e.g. non-learning ChatGPT Team, potentially subject to fees/restrictions), or with open versions of learning AI tools or institutional tool/platform development provided they operate in a secure, user-controlled environment (self-hosted on-premises or externally hosted solely for Participating Institutions or Authorized Users). AI-generated results or platforms may be published for research and teaching provided they do not contain or reproduce closed or CC-BY-NC-ND content (linking is permitted), while commercial use of such AI-generated content or platforms is prohibited.

EPFL Library guidance on scientific information for manuscripts intended for publication states that Large Language Models cannot be listed as authors or co-authors because they lack the ability to be accountable for the content they generate, citing the Swiss Copyright Act (Chapter 2, Art. 6). This statement represents Library guidance for publication manuscripts rather than an independently verified universal legal rule or campus-wide degree thesis requirement.

EPFL study management regulations provide two formal procedures to oppose a grade or notified failure: an internal request for reassessment at EPFL within 10 days, or an external appeal to the Commission de recours interne des EPF within 30 days. Both deadlines are calculated from reception, at the address given by the student, of the grade or notified failure (transcripts sent by post). If a mark is contested on the grounds of an error in correction or assessment, the responsible teacher is the only person authorized to modify the mark subject to justification to the program direction; if contested for a formal defect, Legal Affairs reviews whether a significant formal defect impacted the result to warrant exam cancellation and repetition. Procedural costs are charged ahead of the procedure and refunded only if the student's claim succeeds. These rules govern academic grade disputes; whether specific appeal pathways or different deadlines apply to AI disciplinary misconduct sanctions is not established on this page and remains an evidence gap.