AI cheating among students more than doubled
Academic integrity is under pressure. Since early 2023, 770 cases of unauthorised AI use have led to suspensions across Swedish higher education. 467 of those suspensions were issued this year alone, alongside 75 warnings since 2023.
Generative tools went mainstream in late 2022. By February 2023, Swedish universities recorded their first warning for AI-assisted cheating. The trend hasn't slowed since-reports and suspensions have climbed as access and awareness have grown.
Where the spike is showing
Stockholm University has the highest count so far, with 77 suspensions and 24 warnings across the past three years. The university points to both its size and deliberate staff awareness as drivers-more students, and more informed teachers who know what to look for.
Linköping University reports 63 suspensions related to unauthorised AI use over the past three years, 47 of them this year, plus 12 warnings. Leadership there notes a double effect: more students using AI where it isn't allowed, and teachers getting sharper at explaining boundaries-and reporting violations.
How institutions are responding
Out of 32 universities and colleges that answered questions about exam formats, 27 say they've already changed methods and procedures. Others are in active discussions about how to adapt.
Interestingly, several arts-focused institutions-Beckmans College of Design, Konstfack, the Royal Institute of Art, the Royal College of Music, and Stockholm University of the Arts-report no AI-related cheating cases. Their closer, practice-heavy assessments and ongoing student contact likely play a role. Leaders in the arts stress a different focus: equip students to use AI in ways that genuinely serve their craft.
"AI-proofing" education in practice
Linköping University has rolled out an action plan to review assessment methods across all courses and programmes. The plan funds departmental "AI ambassadors" with SEK 4 million per year for the next two years.
The intent isn't just to curb misconduct. It's to modernise assessment and course content so students graduate ready for workplaces where generative AI is routine.
Policy and penalties at a glance
Under the Higher Education Ordinance, students who use unauthorised aids or attempt to mislead in exams can be suspended for up to six months via a disciplinary board decision. In other situations, the vice-chancellor may issue a warning, which can be appealed to the board.
Read the Higher Education Ordinance.
What educators can do now
- Set explicit AI rules per assignment. State what's allowed, what isn't, and require disclosure of tools, prompts, and assistance received.
- Redesign assessments for process, not just product: in-class writing, oral defenses, practicals, and iterative drafts with version history.
- Use case-based tasks tied tightly to course material. Randomise data or variables, and require personal reflection or decision rationales.
- Blend individual and group work. Assign unique datasets or briefs and add checkpoints that surface the student's own thinking.
- Teach responsible AI use: evaluation of outputs, bias and limits, prompt craft, verification, and proper citation of AI assistance.
- Build capability across staff. Run short workshops, share exemplars, and appoint departmental AI contacts or "ambassadors."
- Detect with care. Treat AI detectors as signals, not verdicts. Triangulate with drafts, interviews, and source checks before escalating.
- Update policies and student handbooks. Make expectations, allowed use cases, and reporting/appeals paths unmistakably clear.
Why the curve may bend back down
An upward curve doesn't have to be the new normal. Clear guidance, thoughtful assessment design, and staff development can reduce misconduct while teaching students how to use AI responsibly. The goal is simple: preserve integrity and improve learning at the same time.
Resources
Your membership also unlocks:
