Description: In February 2025, Tromsø Municipality presented a report proposing the closure of schools and kindergartens. Residents later discovered the report cited numerous non-existent sources. Officials admitted using generative AI to help draft the report, leading to fabricated references. The incident led to public outcry, national commentary, and suspension of the closure plan. It is considered Norway’s first major public-sector AI scandal.
Editor Notes: A couple of notes on two figures who appear in the reporting: Stig Tore Johnsen is the kommunedirektør (municipal chief executive) of Tromsø Municipality. While he publicly accepted responsibility for the inclusion of AI-generated content in the school closure report, there is no indication that he personally deployed the AI system. Thomas Nordahl, a professor whose name was falsely cited in the report as the author of a non-existent book, appears more amused than aggrieved. His public comment suggests concern over the misuse of his name, but not a claim of reputational harm.
Entities
View all entitiesAlleged: Unnamed large language model developer developed an AI system deployed by Tromsø Municipality and Stig Tore Johnsen, which harmed General public of Norway , General public of Tromsø and Thomas Nordahl.
Alleged implicated AI system: Unnamed large language model
Incident Stats
Incident ID
1009
Report Count
2
Incident Date
2025-02-13
Editors
Daniel Atherton
Incident Reports
Reports Timeline

– This proposal will have some consequences, and one of them is a longer travel distance.
This is according to Arne Christian Vangdal, who is leading the project for a new kindergarten and school structure in Tromsø.
– Our main challenge is…
In Tromsø, in northern Norway, the local administration decided it had too many schools and kindergartens for too few children, and they should shut some of them down. So they prepared a very nice 120-page report to support this move. [NRK,…
Variants
A "variant" is an incident that shares the same causative factors, produces similar harms, and involves the same intelligent systems as a known AI incident. Rather than index variants as entirely separate incidents, we list variations of incidents under the first similar incident submitted to the database. Unlike other submission types to the incident database, variants are not required to have reporting in evidence external to the Incident Database. Learn more from the research paper.
Similar Incidents
Did our AI mess up? Flag the unrelated incidents
Similar Incidents
Did our AI mess up? Flag the unrelated incidents