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

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.

Risk classification

• Primary risk domain: 3 Misinformation
• Primary risk subdomain: 3.1 False or misleading information

The AI system generated false and misleading information in the form of fabricated academic citations, which were then published in an official municipal report.

Additional risk subdomains

• 5.1 Overreliance and unsafe use: The municipal administration relied on the AI chatbot to generate factual citations for a public report without verifying the accuracy of the outputs.

Causal factors

• Entity: AI
• Intent: Unintentional
• Timing: Post-deployment

The risk was directly caused by the AI system's hallucination of non-existent academic sources, which was an unintended outcome of using the model.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Limited Risk: The system used was a general-purpose chatbot or text generator used to assist in writing, which falls under the category of chatbots and AI-generated content requiring transparency.

AI system and alleged parties

• AI system: None named
• AI purpose: Writing Assistant; Technical Text Generation
• Behaviour type: Assistant
• Alleged developer: Unknown large language model developers
• Alleged deployer: Tromsø Municipality, Stig Tore Johnsen
• Alleged harmed parties: Thomas Nordahl, General public of Tromsø, General public of Norway

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Minor. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Negligible
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Negligible, indirect Negligible
• Differential treatment: direct Negligible, indirect Negligible
• Civil rights: direct Negligible, indirect Negligible
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Negligible, indirect Negligible
• Epistemic: direct Minor, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Epistemic

Reported: The report explicitly describes epistemic harm through the fabrication of academic sources and citations by an AI system.

Directly caused: The AI system fabricated 11 out of 18 citations in a municipal report, including inventing books and falsely attributing them to real professors (Nordahl and Haug).

People affected

• Occurrences reported: 1
• People reportedly harmed: 2
• People reportedly exposed: 2000

Potential causes

Management

• Rushed Decision Making: Management pressured staff to quickly justify school closures.
• Poor Leadership Oversight: Leaders failed to monitor how administrative reports were compiled.

Technology

• AI Hallucination of References: The chatbot fabricated fictional academic books and citations.
• Unreliable Fact Retrieval: Generative AI models prioritize fluency over factual accuracy.

Data Inputs

• Biased Prompting to AI: AI was used to find sources supporting a predetermined conclusion.

Human Factors

• Blind Trust in AI Outputs: Staff assumed the generated bibliography was accurate without checking.
• Lack of AI Literacy: Users did not understand that LLMs frequently hallucinate sources.

Process and Methods

• No Quality Assurance Process: The municipality lacked a workflow to verify report citations.
• Inadequate Fact Checking: Fictional sources were published because nobody cross-checked them.

Regulatory Environment

• Missing AI Regulations: No clear public sector guidelines existed for using AI in reports.

Information quality

• Classification confidence: High
• Reason for confidence: The reports clearly outline the incident, the specific AI failure (hallucinated citations), the actors involved (Tromso municipality, the professors), and the immediate consequences (suspension of the school closure plan). There is no conflicting information regarding the core facts.
• Ambiguities identified: The exact AI model used is not specified, though it is referred to generally as ChatGPT or a similar chatbot.

Tromso Municipality in Norway used generative AI to draft a policy report, which hallucinated academic citations. This led to public distrust and a temporary suspension of local school closure hearings. The incident represents a localized administrative failure with minor impact on government functions and negligible national security implications.

• Overall national security impact: Minor
• Scope: Single nation
• Primary target: Norway
• Alleged perpetrator: Tromso Municipality

Threat characteristics

• Imminence: Long-term. Represents an ongoing administrative capability and policy concern regarding AI adoption in public services, not an urgent national crisis.
• Autonomy: Human-controlled. The AI system functioned as a drafting assistant under human control, though municipal officials failed to verify the generated output before publication.
• Novelty: Established threat. Generative AI hallucinations and the fabrication of academic citations are well-documented and established technical limitations of large language models.

Impact by dimension

• Physical security: Negligible. The incident involved a written administrative report and posed no threat to physical systems, critical infrastructure, or human safety.
• Information security: Negligible. No hostile foreign information operations or intelligence compromises were involved; the fabricated sources were an unintended system hallucination.
• Sovereignty: Minor. Minor impact on local government functions, resulting in a temporary six-month suspension of a school closure hearing due to public distrust over the flawed report.
• Economic security: Negligible. No threats to strategic industries, financial systems, or national technological competitive advantages were identified.
• Societal stability: Negligible. While causing localized public dissatisfaction and distrust in municipal decisions, the incident did not threaten broad societal stability or human rights.