AIPolicyTracker
Menu

AI incident ·

Purportedly AI-Assisted Report by Tromsø Officials Allegedly Cited Non-Existent Sources in School Closure Proposal

4 news reports Snapshot 7 Sep 2026

In brief

An AI system built by Large Language Model Developers and deployed by Tromso Municipality and Stig Tore Johnsen allegedly harmed Thomas Nordahl, General Public Of Tromso and 6 others.

Risk domain
Misinformation False or misleading information
Occurred
Coverage
4 reportsFeb 2025 - Mar 2025

What happened

In February 2025, Tromsø Municipality reportedly presented a report proposing the closure of schools and kindergartens. Residents reportedly later discovered the report cited numerous non-existent sources. Officials reportedly admitted using generative AI to help draft the report, leading to fabricated references. The incident reportedly led to public outcry and suspension of the closure plan. It is purportedly Norway's first major public-sector AI scandal.

Laws that address this harm

Policy angle: Classified under Misinformation (False or misleading information) in the MIT AI Risk Repository taxonomy; 5 recorded instruments address this use case.

Matched from the record's risk domain and country to the instruments recorded here. A reviewer can correct the match in the repository (data/external/incident_overrides.yaml).

News reports (4)

Titles link to the original publisher; report text is not reproduced here.

  1. Foreslår å legge ned flere barnehager og skoler i Tromsø: – Burde ha skjedd for lenge siden
    nrk.no · Tove Jensen
  2. Norwegian city used AI to generate report that recommended closing schools
    pivot-to-ai.com · David Gerard
  3. Skandalen i Tromsø: Kommunen brukte kilder som KI hadde diktet opp
    digi.no · NTB
  4. Skoleskandalen i Tromsø: Kommunen brukte kilder som KI hadde diktet opp
    utdanningsnytt.no · NTB

Who was involved

Alleged harmed party
Thomas Nordahl, General Public Of Tromso, General Public Of Norway, Students, Teachers, Minors, Educational Communities, Epistemic Integrity

Classification (MIT AI Risk Repository taxonomy)

Risk domain
Misinformation
Causal entity
Human
Intent
Unintentional
Timing
Post-deployment
Harm level
—
Sectors
—
Countries
—

Risk entries describing this failure mode

Entries from the MIT AI Risk Repository coded to subdomain 3.1.

  • Pursuing Consistent Context

    "LLMs have been demonstrated to pursue consistent context [129]–[132], which may lead to erroneous generation when the prefixes contain false information. Typical examples include sycophancy [129], [130], false demonstra...

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Defective Decoding Process

    In general, LLMs employ the Transformer architecture [32] and generate content in an autoregressive manner, where the prediction of the next token is conditioned on the previously generated token sequence. Such a scheme...

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Noisy Training Data

    "Another important source of hallucinations is the noise in training data, which introduces errors in the knowledge stored in model parameters [111]–[113]. Generally, the training data inherently harbors misinformation....

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Hallucinations

    "LLMs generate nonsensical, untruthful, and factual incorrect content"

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Faithfulness Errors

    "The LLM-generated content could contain inaccurate information" which is is not true to the source material or input used

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Factuality Errors

    "The LLM-generated content could contain inaccurate information" which is factually incorrect

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Untruthful Content

    "The LLM-generated content could contain inaccurate information"

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Knowledge Gaps

    "Since the training corpora of LLMs can not contain all possible world knowledge [114]–[119], and it is challenging for LLMs to grasp the long-tail knowledge within their training data [120], [121], LLMs inherently posse...

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

Incidents in the same risk subdomain

All incidents in this subdomain

Source record: incident #1009 on the AI Incident Database · all 4 reports