LATIDIA · Ciberseguridad

Sobreactuación: medición y mitigación de la sobreautorización proactiva en agentes de llamadas de herramientas LLM

arXiv:2610.01508v1 Tipo de anuncio: nuevo Resumen: los agentes de LLM con capacidades de llamada a herramientas pueden acceder a servicios externos y datos privados del usuario, pero pueden recuperar más información de la que solicita explícitamente un usuario

La noticia

arXiv:2610.01508v1 Announce Type: new Abstract: LLM agents with tool-calling capabilities can access external services and private user data, but they may retrieve more information than a user's request explicitly requires. We study this behavior in structured tool-calling agents and term it proactive over-authorization. This setting differs from filesystem-level coding agents because the main risk is unnecessary access to private data. We introduce OverAct, a controlled benchmark spanning eight privacy-sensitive domains with deterministic, judge-free scoring, together with an interpretive decision-theoretic framework that yields three testable predictions. Across seven models from four families, all models significantly exceed authorized scope. Request specificity is the strongest predictor of severity, over-authorization grows sublinearly with tool-pool