Metraipolis › archive › dc_8db2a538d2f8
Where Privacy Risk Lives in English-Source Multilingual RAG: A Stage-Decomposed Audit Across Five Query Languages — arXiv:2608.05163
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.05163
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-25T21:27:01.982Z
- SHA-256 of the archived text
e88a1059a48c16cb3fdfffeff6a8d125a6dc1caf0ce259561bf49002f4e46039- Stable URL of this page
https://metraipolis.com/documento/dc_8db2a538d2f8
How to cite
Where Privacy Risk Lives in English-Source Multilingual RAG: A Stage-Decomposed Audit Across Five Query Languages — arXiv:2608.05163 — arXiv (en), https://arxiv.org/abs/2608.05163 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_8db2a538d2f8 (Metraipolis archive dc_8db2a538d2f8, sha256 e88a1059a48c16cb3fdfffeff6a8d125a6dc1caf0ce259561bf49002f4e46039)
BibTeX
@misc{dc8db2a538d2f8,
title = {Where Privacy Risk Lives in English-Source Multilingual RAG: A Stage-Decomposed Audit Across Five Query Languages — arXiv:2608.05163},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.05163},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_8db2a538d2f8; sha256 e88a1059a48c16cb3fdfffeff6a8d125a6dc1caf0ce259561bf49002f4e46039},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.05163. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
A common assumption holds that switching to a non-English language makes a multilingual RAG system easier to attack for personal information. We test this on an English-source synthetic-PII corpus with five query languages and a two-stage defence (LLM input judge + regex output filter), in a pipeline whose translator, judge, back-translator, and generator are all Qwen2.5-7B -- so every finding below is pipeline-conditional, not a causal ranking of language-inherent risk. Under output-only filtering, English has the highest observed unstructured-PII leak rate; only English-vs-Swahili separates cleanly under document-level bootstrap intervals. Once the input judge is added, residual leaks remain on Arabic and Swahili, and back-translating the query does not close the gap (an ablation we report but cannot use as a causal diagnostic, since the back-translator is also Qwen). On a separate n=17 multilingual-prompted-judge residual corner, attaching the gold corpus document to the input judge blocks 15/17 residual cells. We frame this last result as a mechanism diagnostic, not a deployable defence: it uses oracle retrieval, BLOCK/ALLOW rates are measured on adversarial queries only, and we measure no benign-query false-positive rate and no answer-utility cost. The supplementary material contains code, corpora, queries, and per-trial JSONLs; the priority follow-up is an independent-MT plus non-Qwen-judge replication with a native-speaker query set, scoped in the Limitations section. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-07 in cs.CL). Categorie arXiv: Computation and Language (cs.CL) ; Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2608.05163 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-25T21:24:38.172Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-08-07?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).