This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
An Estonian court fined ten plaintiffs after they submitted a legal filing containing false academic references and non-existent researchers generated by an artificial intelligence tool.
Ten plaintiffs challenging a logging permit in Estonia reportedly received fines after submitting an AI-assisted document citing researchers who did not exist and studies the court could not locate. Tallinn Administrative Court held them responsible for checking the references. The plaintiffs denied intending to mislead the court; the fines became final after appeals.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI systems that inadvertently generate or spread incorrect or deceptive information, which can lead to inaccurate beliefs in users and undermine their autonomy. Humans that make decisions based on false beliefs can experience physical, emotional or material harms
Also classified as
The AI system failed to perform reliably by hallucinating non-existent facts and citations, demonstrating a lack of capability and robustness in generating factual information.
The AI system generated false academic references and non-existent researchers, which constitutes the generation of false or misleading information that was subsequently submitted to a court.
AI system
Due to a decision or action made by an AI system
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The risk was caused by the AI system hallucinating false academic citations, which was an unexpected and unintentional outcome of the plaintiffs using the tool to draft their legal document.
Added categories peak at 2: Minor (Epistemic Harm, Direct)
Ten plaintiffs in Estonia were fined after submitting an AI-generated legal filing containing fabricated academic references regarding a logging permit. This incident highlights minor, manageable disruptions to judicial proceedings from unverified AI hallucinations, but poses no significant threat to national security.
The AI system acted as an assistant to help the plaintiffs draft their court documents and identify arguments, but required human oversight to validate the accuracy of the generated text.
Limited Risk: The system used was a general-purpose AI tool/chatbot capable of generating text and legal references, which falls under transparency obligations to ensure users are aware of its nature.
The AI system was used by the plaintiffs to draft court documents and help non-lawyers identify arguments and write legal filings.
The reports provide consistent, detailed, and clear factual accounts of the court proceedings, the specific fines levied, the nature of the AI's hallucination, and statements from the court chair. There are no conflicting accounts or major ambiguities regarding the core event.
None
Technology
AI Hallucination
The AI tool generated completely fictional Estonian researchers and studies.
Unreliable Legal Advice
The AI tool lacked reliability and quality as a source for legal citations.
Plausible-Looking Output
AI-generated text looked convincing enough to deceive uncritical users.
Data Inputs
Unverified AI Sources
The data fed into the court filing was not verified against original sources.
Human Factors
Lack of Source Criticism
Plaintiffs were not critical enough of the sources generated by the AI.
Gross Negligence
Plaintiffs submitted false academic references without checking if they existed.
Misplaced Trust in AI
Plaintiffs assumed AI use was safe because its use is not explicitly banned.
Process and Methods
Absence of Verification Step
No process was followed by the plaintiffs to cross-check AI references.
Use of AI for Court Filings
Using AI to draft formal legal documents without professional oversight.
Regulatory Environment
Permissive AI Rules
Estonian court rules allow AI use but punish submitting false information.
AI system involved
An AI system was involved, but the reports do not identify it.
Identified by automated extraction; can differ from the alleged developers above.
