A student at the University of South-Eastern Norway was suspended for one year after submitting AI-generated deepfake videos for Spanish coursework. Faculty identified the fraud due to unnatural facial movements, inconsistent language ability, and a lack of synchronization between speech and mouth movements. The university is now considering changing its examination methods to prevent similar incidents of AI-enabled academic dishonesty.
A student at the University of South-Eastern Norway reportedly submitted multiple deepfake videos for Spanish coursework, featuring an AI-generated likeness and synthetic voice. Faculty reportedly noted unnatural facial movements and inconsistent language ability. The case was reportedly ruled academic misconduct, resulting in annulled grades and a year-long suspension, marking one of Norway's first confirmed deepfake-related cheating cases.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Using AI systems to gain a personal advantage over others such as through cheating, fraud, scams, blackmail or targeted manipulation of beliefs or behavior. Examples include AI-facilitated plagiarism for research or education, impersonating a trusted or fake individual for illegitimate financial benefit, or creating humiliating or sexual imagery.
The student used AI-generated deepfake videos to commit academic fraud and cheat on coursework to gain an unfair academic advantage.
Human
Due to a decision or action made by humans
Intentional
Due to an expected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The academic fraud was caused by a human student intentionally using post-deployment AI tools to generate fake coursework videos.
Added categories peak at 2: Minor (Epistemic Harm, Inferred)
A student at a Norwegian university used AI-generated deepfake videos to commit academic fraud in a Spanish course. The incident is a localized academic disciplinary matter with negligible national security implications.
The student used deepfake generation software as a tool to perform a specific function (generating synthetic video and audio) without autonomous planning or agency.
Limited Risk: The report describes the use of AI-generated content (deepfakes), which is classified under Risk Level 3 due to transparency and disclosure requirements.
The student used AI tools to generate deepfake videos of herself speaking Spanish, manipulating both her face and voice to pass coursework.
The report clearly outlines the facts of the cheating incident, the detection method, the university's response, and the student's admission. There is little ambiguity about what transpired.
None.
Technology
Accessible Deepfake Tools
Free and cheap AI tools make video and voice manipulation easy for anyone.
Realistic AI Generation
AI can generate realistic voices and text, making detection challenging.
Data Inputs
Pre-recorded Video Submissions
Using pre-recorded files allows students to manipulate and edit media.
Lack of In-person Verification Data
Online course format meant teachers had no baseline of student's real speech.
Human Factors
Student Intent to Cheat
The student chose to use AI to bypass actual language learning and exams.
Lack of Student Engagement
Student did not attend sessions or respond to the university's inquiries.
Process and Methods
Vulnerable Evaluation Format
Relying on pre-recorded videos is naive in the era of advanced AI tools.
Difficulty Proving AI Use
Proving academic dishonesty with AI is highly challenging for educators.
Management
Unpreparedness for AI Cheating
Management failed to anticipate and prepare for AI-manipulated media.
Lack of Proactive Risk Assessment
No proactive assessment of AI risks on remote oral examination methods.
Regulatory Environment
Outdated Academic Integrity Rules
University regulations were not prepared for AI-generated video submissions.