Why this policy?

AI tools may be used at the College where they support a student's learning, but not where they replace it. Four principles apply: the work must respect academic integrity, the use of AI must be compatible with what the course work is designed to teach, that use must be openly acknowledged, and students take full responsibility for the content they submit. Outputs are to be checked for accuracy, and data protection and intellectual property rules apply as anywhere else.

Key principle: Learning outcomes come first

Every course is built around intended learning outcomes, that is, what a student is expected to have achieved upon completion of all deliverables. If AI use prevents students from achieving the intended outcomes of an assignment, it is not conducive to the successful completion of the course. If AI effectively supports their learning, it offers – within the permitted scope – new, valuable possibilities.

In consultation with their department, professors therefore set the permitted level of generative AI for each assignment in light of the competences it is meant to develop and assess. The permitted level is indicated in the ECTS card and course outline, and may differ between assignments of the same course.

Non-generative use of AI vs. generative use of AI

At the College, we differentiate between non-generative use of AI and generative use of AI. We do this to acknowledge that many standard tools our students have used for years are now AI-powered.

• Non-generative use → AI-powered support functions that do not create content (e.g. reference managers, searching for information, spell and grammar checks). Such AI use is allowed by default, except in language courses. It ends where a tool rewrites, reformulates or translates sentences, or produces new content.
• Generative use → use of (usually prompt-based or agentic) tools to create, analyse, revise, machine translate or adapt content. Allowed only within the category specified for the assignment (explained below).

Four categories of generative AI use

Each outline and ECTS card of a course indicates which category of generative use of AI applies – either for all assignments of the course or with a differentiation per assignment. The category is a ceiling, not an expectation: students are never required to use AI.

Declarations and disclosure

Transparency is the core of the framework. Based on Art. 38 of the Study Regulations, common reporting obligations approved by the Academic Council apply across all departments. What students submit depends on the permitted category:

• Where generative AI is prohibited, students confirm compliance with the rules via a Declaration of Honour.
• Where AI use is permitted (restricted, bounded or open) and students made use of it, they submit an AI Use Statement: which tools were used and for what purpose, which outputs were included and how they were integrated or modified, and the degree of reliance on AI in producing the final work.
• For the Master's thesis, the Statutory Declaration is accompanied by the standardised AI Disclosure Form.

Note: At the College, we do not treat AI tools in themselves like a source. There is per se no requirement to list AI use in the bibliography. Under the open category, professors or departments may require footnote-style or in-text disclosure of specific AI contributions.

If questions arise about a student's work

• A conversation comes first. Where questions arise, the aim is to understand how the student worked and to make expectations clear. Students are always invited to a hearing with a panel consisting at least of a professor and an Academic Assistant (in line with Art. 41 of the Study Regulations), where they can explain their way of working and their use of AI. A hearing may well conclude with guidance on future assignments; such guidance is not a sanction.
• Sanctions require evidence. A sanction is only ever imposed after a thorough examination of the case, a hearing, and sufficient evidence of a transgression.
• Decisions on possible sanctions are taken at departmental level or – depending on the nature of the case – at Academic Council level.
• Sanctions, if applied, are progressive and proportionate to the scope and severity of the transgression. The College distinguishes minor, significant and major transgressions. Consequences range from a deduction of points to a mark of 0 for the assignment or the course as a whole and, in cases of systematic academic misconduct, expulsion from the programme.

Plagiarism, falsification of data, etc. are sanctioned independently and additionally to any sanction for unauthorised AI use. Students remain responsible for all parts of their submitted assignment, including any error, invented source or bias introduced by an AI tool.

Support & resources

• Professors and Academic Assistants of the department → first point of contact when unsure.
• Library → resources on AI tools and their use in academia.

Key takeaways

• The College's AI rules are based on a learning outcomes-based framework.
• Professors decide, in consultation with their department, on the category of permitted generative AI use per assignment/course. By default, the use of AI for the Master's thesis is restricted, and prohibited in exams.
• Acknowledgement and transparency are key. Students are always heard before any sanction.
• Above all: the framework is here to guide the learning process and strengthen academic integrity, and to help our students achieve the knowledge, skills and competences that they came to acquire at the College of Europe and that will subsequently help them in their professional career.