Related papers: Do LLMs Fabricate Legal Citations? A Bilingual Benchmark on Saudi Data Protection Law and the GDPR

Do LLMs Fabricate Legal Citations? A Bilingual Benchmark on Saudi Data Protection Law and the GDPR

URL: http://arxiv.org/abs/2607.11127v1
Date: Mon, 13 Jul 2026 06:03:54 GMT
Title: Do LLMs Fabricate Legal Citations? A Bilingual Benchmark on Saudi Data Protection Law and the GDPR
Abstract summary: A wrong statutory citation can silently propagate into legal advice, compliance documentation, and policy decisions.<n>We introduce a benchmark of 120 questions probing whether freely accessible language models (LLMs) fabricate article citations.<n>The benchmark pairs direct citation retrieval questions with false premise verification probes and deliberately unanswerable "trap" questions.
Score: 0.0
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Abstract: Organizations and regulators increasingly consult large language models (LLMs) for regulatory-compliance questions, yet a wrong statutory citation can silently propagate into legal advice, compliance documentation, and policy decisions. We introduce a bilingual benchmark of 120 questions probing whether freely accessible LLMs fabricate article citations for two data-protection instruments: the EU General Data Protection Regulation (GDPR) and the Saudi Personal Data Protection Law (PDPL). The benchmark pairs direct citation retrieval questions with false premise verification probes and deliberately unanswerable "trap" questions -- including questions about a repealed article and about deadlines that exist only in implementing regulations, not in the law itself. Every question is posed in both Arabic and English, and all scoring is fully automatic against a manually verified gold reference. Evaluating three freely accessible models (Gemini 2.5 Flash, GPT-OSS-120B, Nemotron-3-Super-120B), we find a dramatic jurisdiction gap: near-ceiling citation accuracy on the GDPR (94-100% on direct retrieval) against majority fabrication on the Saudi PDPL (60-77%), invariant to query language; the highest fabrication rates (67%) arise from statute-vs-regulations confusion, and 91% of fabricated citations are asserted with confidence >= 0.8. Fabrication tracks the jurisdiction of the law, not the language of the query, and model confidence provides no protection -- indicating that verbatim-verification safeguards, rather than model self confidence, must gate any institutional reliance on LLMs for compliance screening.

Related papers

TW-LegalBench: Measuring Taiwanese Legal Understanding [1.8953545128709688]
Large language models (LLMs) have shown impressive capabilities across diverse tasks, yet their performance on jurisdiction-specific legal reasoning remains underexplored.<n>We present TW-LegalBench that utilizes Taiwanese legal system's rich official corpus open to the public to fill the gap in evaluating LLMs on Taiwanese law.
arXiv Detail & Related papers (2026-06-17T05:25:21Z)
LegalCiteBench: Evaluating Citation Reliability in Legal Language Models [14.281332347684872]
LegalCiteBench is a benchmark for studying closed-book citation recovery, citation verification, and case matching in legal language models.<n>The benchmark covers five citation-centric tasks: citation retrieval, citation completion, citation error detection, case matching, and case verification and correction.
arXiv Detail & Related papers (2026-05-11T08:37:54Z)
ContextLens: Modeling Imperfect Privacy and Safety Context for Legal Compliance [49.524070843587594]
We propose ContextLens, a semi-rule-based framework that leverages large language models (LLMs) to ground input context in the legal domain.<n>We conduct experiments on existing compliance benchmarks that cover the General Data Protection Regulation () and the EU AI Act.<n>The results suggest that our ContextLens can significantly improve LLMs' compliance assessment and surpass existing baselines without any training.
arXiv Detail & Related papers (2026-04-14T05:35:38Z)
Towards Reliable Truth-Aligned Uncertainty Estimation in Large Language Models [54.51264434040939]
Uncertainty estimation (UE) aims to detect hallucinated outputs of large language models (LLMs)<n>We formalise this phenomenon as proxy failure, since most UE metrics originate from model behaviour.<n>We propose Truth AnChoring (TAC), a post-hoc calibration method to remedy UE metrics.
arXiv Detail & Related papers (2026-04-01T03:42:16Z)
Towards Comprehensive Stage-wise Benchmarking of Large Language Models in Fact-Checking [64.97768177044355]
Large Language Models (LLMs) are increasingly deployed in real-world fact-checking systems.<n>We present FactArena, a fully automated arena-style evaluation framework.<n>Our analyses reveal significant discrepancies between static claim-verification accuracy and end-to-end fact-checking competence.
arXiv Detail & Related papers (2026-01-06T02:51:56Z)
EU-Agent-Bench: Measuring Illegal Behavior of LLM Agents Under EU Law [39.146761527401424]
EU-Agent-Bench is a verifiable benchmark that evaluates an agent's alignment with EU legal norms.<n>Our benchmark spans scenarios across several categories, including data protection, bias/discrimination, and scientific integrity.<n>We release a public preview set for the research community, while holding out a private test set to prevent data contamination.
arXiv Detail & Related papers (2025-10-24T14:48:10Z)
ClauseLens: Clause-Grounded, CVaR-Constrained Reinforcement Learning for Trustworthy Reinsurance Pricing [0.0]
ClauseLens is a clause-grounded reinforcement learning framework for treaty quoting.<n>It produces transparent, regulation-compliant, and risk-aware treaty quotes.<n>It reduces solvency violations by 51%, improves tail-risk performance by 27.9%, and achieves 88.2% accuracy in clause-grounded explanations.
arXiv Detail & Related papers (2025-10-09T16:43:49Z)
Towards Automated Regulatory Compliance Verification in Financial Auditing with Large Language Models [2.502173772742615]
We probe the efficiency of publicly available Large Language Models (LLMs) in the realm of regulatory compliance across different model configurations.<n>We find that the open-source Llama-2 70 billion model demonstrates outstanding performance in detecting non-compliance or true negative occurrences.<n> proprietary models such as GPT-4 perform the best in a broad variety of scenarios, particularly in non-English contexts.
arXiv Detail & Related papers (2025-07-22T14:39:54Z)
Context Reasoner: Incentivizing Reasoning Capability for Contextualized Privacy and Safety Compliance via Reinforcement Learning [53.92712851223158]
We formulate safety and privacy issues into contextualized compliance problems following the Contextual Integrity (CI) theory.<n>Under the CI framework, we align our model with three critical regulatory standards: EU AI Act, and HIPAA.<n>We employ reinforcement learning (RL) with a rule-based reward to incentivize contextual reasoning capabilities while enhancing compliance with safety and privacy norms.
arXiv Detail & Related papers (2025-05-20T16:40:09Z)
LEXam: Benchmarking Legal Reasoning on 340 Law Exams [76.3521146499006]
We introduce textscLEXam, a novel benchmark derived from 340 law exams spanning 116 law school courses across a range of subjects and degree levels.<n>The dataset comprises 4,886 law exam questions in English and German, including 2,841 long-form, open-ended questions and 2,045 multiple-choice questions.<n>Our results underscore the effectiveness of the dataset in differentiating between models with varying capabilities.
arXiv Detail & Related papers (2025-05-19T08:48:12Z)
InternLM-Law: An Open Source Chinese Legal Large Language Model [72.2589401309848]
InternLM-Law is a specialized LLM tailored for addressing diverse legal queries related to Chinese laws. We meticulously construct a dataset in the Chinese legal domain, encompassing over 1 million queries. InternLM-Law achieves the highest average performance on LawBench, outperforming state-of-the-art models, including GPT-4, on 13 out of 20 subtasks.
arXiv Detail & Related papers (2024-06-21T06:19:03Z)

This list is automatically generated from the titles and abstracts of the papers in this site.