Skip to main content
NOPE
Medium reported severity

Leiden law student's thesis invalidated and exams barred over references he says an AI tool distorted or invented; Council of State annuls fraud sanction

A law student at Leiden University used a generative AI tool to spell-check his bachelor thesis and to check and order its reference list. Seven of the thesis's footnote references (one of them appearing twice, out of 36 footnotes with references) were wrong in the final version; according to his explanation, the tool had distorted some of them, for example with wrong years or author initials, and invented others. On 29 January 2026 the faculty's exam board declared the thesis invalid and excluded him until 28 January 2027 from the thesis track and every exam in the faculty's programmes, reasoning that the fictitious references indicated unauthorised use of generative AI and therefore fraud. The university's appeals board for examinations found in March 2026 that he had plausibly shown he used generative AI to prepare his reference list, which it called a permitted use, but held him responsible for the non-existent sources; the exam board then re-imposed the invalidation with an exclusion until 29 July 2026, and the appeals board upheld that in July. On 14 October 2026 the Administrative Jurisdiction Division of the Council of State (Raad van State) found his explanation plausible, held that no more than carelessness had been shown, and annulled the sanction; the thesis must now be graded, and the examiner may take the incorrect references into account in the grade.

AI system
Unidentified AI writing tool
Occurred
29 Jan 2026
Reported
14 October 2026
Event location
Netherlands
What the AI did
Communicated with the person
Reported harm
Educational Harm
Whose AI use
Their own AI use
Setting
Education · Justice
Evidence
AI involvement supported · Causal attribution supported · 2 sources, 1 underlying account
7 claims: 6 documented, 1 reported. 3 open questions
People reported harmed
1 person

AI system as recorded: An unnamed generative AI tool that the student used to spell-check his bachelor thesis and to check and order its reference list, according to the Council of State ruling

What Happened

The thesis. The student, enrolled in Leiden University's bachelor programme in law, used AI for a spelling check of his bachelor thesis and to check and order its reference list, according to the ruling. In the final version seven references in footnotes were rendered incorrectly, one of them twice, out of 36 footnotes with references. In a written response of 28 January 2026, made right after the examiner found that not all sources were correct, he set out which references the AI had distorted (wrong years or author initials, for example) or had made up, and what the correct references should have been. The correct sources appeared in shortened form in an earlier version of the thesis written before he used AI.

The sanction. On 29 January 2026 the exam board of the Faculty of Law declared the thesis invalid and excluded him until 28 January 2027 from the bachelor thesis track and any exam in a programme taught by the faculty. The board said the fictitious references indicated unauthorised use of generative AI and therefore fraud, and considered the sanction appropriate because he had twice before been found to have committed fraud. On 4 March 2026 the university's appeals board for examinations (CBE) upheld his administrative appeal: it found that he had plausibly shown he used generative AI to compile his reference list, a permitted form of AI use, that the non-existent sources were nonetheless his responsibility, and that the sanction had disproportionately heavy consequences; with no fraud found in the main text, it considered a half-year exclusion more fitting. On 6 May 2026 the exam board again declared the thesis invalid and excluded him until 29 July 2026, and on 8 July 2026 the CBE dismissed his appeal against that decision.

The ruling. After a hearing on 3 September 2026, the Administrative Jurisdiction Division of the Council of State ruled on 14 October 2026 that not every incorrect reference is fraud and that the scale and seriousness of the errors and the student's explanation matter. It found his explanation plausible, noted that he was at fault for not checking all the references himself after using AI, and held that the CBE had not made it plausible that there was more than carelessness, so fraud was not established beyond reasonable doubt. It annulled the decisions of 8 July and 6 May 2026 and ordered that the thesis still be assessed; the examiner may decide whether the carelessness affects the grade. The ruling does not name the AI tool and does not say whether the exclusion kept him from any particular exam.

Reported harm

The exam board declared the student's thesis invalid and excluded him from the thesis track and the faculty's exams, first until January 2027 and then until 29 July 2026, over references that according to his accepted explanation an AI tool had distorted or invented; the Council of State annulled the sanction on 14 October 2026.

How Harm Occurred

A generative AI tool used to order a thesis's reference list returned distorted and invented references, according to the student's explanation; the exam board read the non-existent references as evidence of unauthorised AI use and fraud and invalidated the thesis and excluded him from exams until the Council of State annulled the sanction.

Outcome

Resolved

The Administrative Jurisdiction Division of the Council of State (case 202602163/1/A2, ruling of 14 October 2026, ECLI:NL:RVS:2026:6088) upheld the appeal, annulled the appeals board's decision of 8 July 2026 and the exam board's decision of 6 May 2026, ordered the thesis to be assessed in light of the ruling and left it to the examiner whether the carelessness affects the grade; the university bodies must pay the student's costs.

What remains unknown

  • Which AI tool the student used; the ruling does not name it.
  • Whether the exclusion orders kept him from any particular exam or delayed his studies, and what grade the thesis eventually received.
  • The exact wording of the incorrect references and how the examiner discovered them.

What the evidence supports

AI involvement: supported. The Council of State ruling records the student's explanation that the AI tool he used to prepare his footnote references did not carry over all sources from his draft correctly and that it distorted some references (wrong years or author initials) and invented others; the university's appeals board found he had plausibly shown he used generative AI for his reference list, and the Council of State found his explanation plausible, noting that the correct sources were in an earlier version written before he used AI. The exam board itself attributed the fictitious references to generative AI and sanctioned him for them. The AI tool's output, returned to the student and kept in his thesis, is what the sanction was based on.

7 claims: 6 documented, 1 reported. What the statuses mean

Documented On 29 January 2026 the Leiden law faculty's exam board declared the student's bachelor thesis invalid and excluded him until 28 January 2027 from the thesis track and any exam in the faculty's programmes, because, in its view, non-existent sources in the references indicated unauthorised use of generative AI and therefore fraud.

Causal attribution. The decisions are recorded in the ruling; the PONT summary restates them.

  • data.rechtspraak.nl(opens in new tab) supports · Dutch
    'Bij beslissing van 29 januari 2026 heeft de examencommissie de scriptie van'; 'tot en met 28 januari 2027 uitgesloten van deelname aan het bachelorscriptietraject en enig ander tentamen van een door de faculteit Rechtsgeleerdheid verzorgde opleiding'; 'De examencommissie heeft deze sanctie opgelegd omdat in het verwijzingsapparaat van zijn bachelorscriptie een aantal niet bestaande bronnen is opgenomen. Volgens de examencommissie duiden deze fictieve bronvermeldingen op ongeoorloofd gebruik van generatieve AI en dus fraude.'
  • privacy-web.nl(opens in new tab) supports · Dutch
    'De examencommissie beschouwde dit als fraude.'
Documented The student stated that he had not invented sources himself and that the AI tool he used to prepare the references had not carried over all sources from his draft correctly, affecting seven references in footnotes of the final version, one of them twice, out of 36 footnotes with references.

Causal attribution. The ruling documents that he made this argument; the count of errors is stated by the court.

  • data.rechtspraak.nl(opens in new tab) supports · Dutch
    'Hij heeft niet zelf bronnen verzonnen, maar de AI-tool die hij heeft gebruikt bij het opstellen van het verwijzingsapparaat heeft niet alle bronnen uit het concept van zijn scriptie correct overgenomen.'; 'Het gaat om zeven foutieve verwijzingen - waarvan één tweemaal - op een totaal van zesendertig voetnoten met bronverwijzingen.'
Reported He had used AI to spell-check the thesis and to check and order its reference list, and the AI distorted some references (for example wrong years or author initials) and invented others.

Causal attribution. The student's own account; the appeals board and the Council of State found it plausible, but no examination of the tool's output is described.

  • data.rechtspraak.nl(opens in new tab) supports · Dutch
    'heeft toegelicht dat hij voor zijn scriptie AI heeft gebruikt voor een spellingcontrole en om zijn bronnenlijst te controleren en te ordenen'; 'toegelicht welke bronnen door AI zijn verdraaid, zoals verkeerde jaartallen of voorletters van auteurs, of door AI zijn bedacht'
  • privacy-web.nl(opens in new tab) supports · Dutch
    'Sommige bronnen bevatten verkeerde jaartallen of auteursnamen, andere waren door AI verzonnen.'
Documented On 4 March 2026 the university's appeals board for examinations upheld his administrative appeal, finding that he had plausibly shown he used generative AI to compile his reference list, a permitted use, that the non-existent sources were his responsibility, and that a half-year exclusion was a more fitting measure.

Causal attribution. Recorded in the ruling.

  • data.rechtspraak.nl(opens in new tab) supports · Dutch
    'Bij beslissing van 4 maart 2026 heeft het CBE het door'; 'heeft voldoende aannemelijk gemaakt dat hij met het oog op het opmaken van zijn bronnenlijst gebruik heeft gemaakt van generatieve AI. Dat is een geoorloofde vorm van AI-gebruik. Dat daarbij niet-bestaande bronnen in het verwijzingsapparaat zijn terechtgekomen, komt evenwel voor zijn rekening.'; 'voor de duur van een half jaar een passender maatregel'
Documented On 6 May 2026 the exam board again declared the thesis invalid and excluded him until 29 July 2026, and on 8 July 2026 the appeals board dismissed his appeal against that decision.

Causal attribution. Recorded in the ruling.

  • data.rechtspraak.nl(opens in new tab) supports · Dutch
    'Bij beslissing van 6 mei 2026 heeft de examencommissie de scriptie van'; 'tot en met 29 juli 2026 uitgesloten van deelname aan het bachelorscriptietraject en enig ander tentamen'; 'Bij beslissing van 8 juli 2026 heeft het CBE het door'; 'ingestelde administratief beroep ongegrond verklaard'
Documented On 14 October 2026 the Council of State found that the appeals board had not made it plausible that there was more than carelessness, so fraud was not established beyond reasonable doubt; it noted that the correct sources were in an earlier version written before AI was used, annulled the decisions of 8 July and 6 May 2026, ordered that the thesis still be assessed, and left it to the examiner whether the carelessness affects the grade.

Causal attribution. The ruling is the decision itself.

  • data.rechtspraak.nl(opens in new tab) supports · Dutch
    'Datum uitspraak: 14 oktober 2026'; 'al deze juiste bronnen in een eerdere versie van de scriptie, vóór het gebruik van AI, in verkorte vorm zijn terug te vinden'; 'heeft het CBE onvoldoende aannemelijk gemaakt dat sprake is van meer dan alleen een onzorgvuldigheid'; 'alsnog worden beoordeeld, met inachtneming van wat de Afdeling heeft overwogen'; 'Het is aan de examinator of, en, zo ja, welke gevolgen hij in dat kader aan de geconstateerde onzorgvuldigheid verbindt voor het cijfer.'
  • privacy-web.nl(opens in new tab) supports · Dutch
    'De Afdeling bestuursrechtspraak van de Raad van State geeft hem gelijk.'
Documented The Council of State held that the student was at fault for not independently checking all the references again after using AI.

Causal attribution. The court's finding; it bears on the student's own share in the errors.

  • data.rechtspraak.nl(opens in new tab) supports · Dutch
    'te wijten is dat hij na het gebruik van AI niet nogmaals zelfstandig alle bronvermeldingen in de voetnoten heeft gecontroleerd'
  • privacy-web.nl(opens in new tab) supports · Dutch
    'De Raad van State stelt expliciet dat hij de bronvermeldingen na het gebruik van AI zelfstandig had moeten controleren.'

Sources

2 sources inspected, from 1 underlying account. Sources that repeat one account do not corroborate each other.

How the sources were read, and where the events happened

Read in Dutch by the research agent (an AI) on 2026-10-17 and translated by the research agent (an AI) into English; no human translator or reviewer was involved. Full text of the Council of State ruling 202602163/1/A2 of 14 October 2026 (ECLI:NL:RVS:2026:6088) from the rechtspraak.nl open data service, HTTP 200; the student is anonymised in the ruling. Applies to s1.

Read in Dutch by the research agent (an AI) on 2026-10-17 and translated by the research agent (an AI) into English; no human translator or reviewer was involved. PONT Data & Privacy editorial summary of the same ruling, 15 October 2026, HTTP 200, full text read; it derives from the ruling and is not independent of it. Applies to s2.

Event countries: Netherlands. Affected-person countries: Unknown. Court countries: Netherlands.

The sanctioning exam board belongs to the Faculty of Law of Leiden University and the appeal was decided by the Council of State, both in the Netherlands. The ruling withholds the student's place of residence, so that field is left unknown.

Reviewed for publication 2026-10-17: Published on the Council of State's ruling of 14 October 2026: the student's explanation, found plausible by the university appeals board and the court, is that a generative AI tool he used to order his reference list distorted or invented seven references; the exam board treated those references as evidence of AI-enabled fraud, invalidated the thesis and excluded him from exams, and the court annulled the sanction because only carelessness had been shown.

People described

A bachelor law student at Leiden University

People reported harmed in this case

1 person

1 AI participant · 0 other people harmed

One person: the appellant student, who used the AI tool himself (ruling, paragraphs 1 to 6).

Counted once within this case. The same person may appear in other cases. This count does not establish AI causation.

Tags

#academic-integrity#thesis#non-existent-references#fraud-sanction-annulled#netherlands#administrative-court

Cite this case

Compiled per our published methodology: verification statuses, sourcing standards, and corrections process.

APA

NOPE. (2026). Leiden law student's thesis invalidated and exams barred over references he says an AI tool distorted or invented; Council of State annuls fraud sanction. AI incidents. https://nope.net/incidents/2026-leiden-law-student-thesis-sanction-over-ai-invented-references-annulled-by-council-of-state

BibTeX

@misc{2026_leiden_law_student_thesis_sanction_over_ai_invented_references_annulled_by_council_of_state,
  title = {Leiden law student's thesis invalidated and exams barred over references he says an AI tool distorted or invented; Council of State annuls fraud sanction},
  author = {NOPE},
  year = {2026},
  howpublished = {AI incidents},
  url = {https://nope.net/incidents/2026-leiden-law-student-thesis-sanction-over-ai-invented-references-annulled-by-council-of-state}
}

Related cases

Low Unidentified AI writing tool (suspected)

Dutch Council of State upholds Maastricht University fraud sanction on a law student whose assignment cited non-existent sources the university attributed to AI

On 11 December 2025 an examination board at Maastricht University found that a European Law School student had committed fraud in a weekly assignment whose purpose was to find and correctly cite legal sources. The board said three of the four sources contained wrong journals, years, titles and page numbers, that the sources did not exist, and that the way they were included pointed to the use of AI; it concluded the student had made unauthorised use of AI. The sanction was a formal warning in the student's file, invalidation of all assessment components of the course and exclusion from the resit. The student argued that no technical analysis or AI-detection tool had been used, that she was never asked whether she had used AI, and that the errors came from completing the work in a hurry from memory and notes with a poor internet connection. The university's examination appeals board rejected her appeal on 5 March 2026, and on 12 August 2026 the Administrative Jurisdiction Division of the Council of State dismissed her further appeal and her claim for damages for reputational and psychological harm.

High Grok and Stability AI tools (reported)

Tennessee minors v. xAI: three girls say a man one of them knew used an AI app to make nude images and video of them and other minors, at least 19 in all, and traded them online; they sued xAI, and later also Stability AI, over the tool

Three Tennessee girls, two still minors, sued xAI in the Northern District of California on 16 March 2026 over AI-generated nude images and a video made of them as minors. According to the complaint, on or about 6 December 2025 an anonymous Instagram user sent one of them sexual AI images of herself made from her own photos, including her homecoming and yearbook pictures, and a link to a Discord server created by a person she knew, which held images and videos of at least 18 other minor girls, including the two other plaintiffs, many from her school. Local police arrested the man in late December 2025; the other two plaintiffs learned from investigators in February 2026 that he had made nude images of them, and he is alleged to have traded the files on Telegram and Mega. The plaintiffs describe anxiety, sleeplessness, nightmares, self-isolation and needing academic accommodations. Their first complaint said law enforcement linked the app he used to xAI's Grok technology; the July amended complaint says the app relied on Stability AI's tools and names Stability AI as a defendant. The case is at an early stage and no court has ruled on the allegations.

Low Unidentified AI writing tool (suspected)

St Andrews medical student says a marker's suspicion of AI use led to an upheld misconduct finding, overturned at a second hearing after appeal

The BBC reported on 10 August 2026 that a University of St Andrews medical student received an academic misconduct alert in August of the previous year saying his dissertation had been flagged as AI-generated. According to the university, the dissertation was assessed by two human markers, one of whom raised concerns about possible AI use; the marker's notes, seen by the BBC, cite a polished, uniform tone, flawless grammar, formulaic paragraph structure and frequent dashes. The student says the link to his virtual hearing arrived five minutes after the start time. A week after the hearing the allegations were upheld, which capped his mark at a bare pass. The university accepted his appeal because of the late link, and at a second hearing in mid-November it did not uphold the allegations. He later received his full mark. The student says he was consumed by worries throughout that semester, that he deferred his exams and that his graduation was delayed until December. The university says the appeal succeeded solely on a procedural issue. The report describes no AI-detection tool, and the student says he did not use AI.

Low Unidentified AI writing tool (suspected)

A University of Houston-Downtown senior was given an F in a music appreciation course after the professor claimed journal entries and discussion posts were AI-generated; two appeals failed before a student discipline committee overturned the finding

FOX 26 Houston reported on 2 October 2026 that a senior studying business management at the University of Houston-Downtown received an F in a music appreciation course 'last August' after the professor claimed the student had submitted several journal entries and discussion posts 'determined to be copy and pasted AI generated'. The student appealed. The professor and then the department chair denied the appeal, and a student discipline committee of four deans ruled in the student's favour. The station quotes the decision as based on concerns about the reliability and consistency of the evidence used to support the original finding. The report does not say how the work was determined to be AI-generated or whether a detection tool was used. The student described being shocked and incensed by the grade and relieved by the outcome.

If you or someone you know is struggling, free and confidential support is available. Find a helpline near you at Signpost.

Back to all incidents