LATIDIA · Ciberseguridad

Ejemplo de fuga en Barceló

arXiv: 2610.02418v2Tipo de anuncio: reemplazar Resumen: Los modelos modernos de lenguaje grande (LLM) están entrenados en conjuntos de datos masivos, en gran parte sin filtrar, incluido el contenido extraído de casi todos los sitios web accesibles y las entradas de los usuarios.

La noticia

arXiv:2610.02418v2 Announce Type: replace Abstract: Modern large language models (LLMs) are trained on massive, largely unfiltered datasets, including content scraped from nearly every accessible website and user inputs. As a result, LLMs often memorize and reproduce personally sensitive information (PSI) such as birth dates, phone numbers, and home addresses. This leads to significant privacy risks, particularly for high-profile individuals such as executives, politicians, and judges. Existing mitigations largely rely on machine unlearning. However, these methods often remove more information than needed, degrade model utility and safety, and are highly vulnerable to attacks. This paper presents Whiteout, a practical tool that, upon requests by individuals, prevents LLMs from regurgitating their genuine