Pasar al contenido principal

SISTEMAS INFORMATICOS INDUSTRIALES CONFIABLES EN LA ERA DE LA COMPUTACION EN EL CONTINUO

PID2024-155230OB-C43

•
Nombre agencia financiadora Agencia Estatal de Investigación
Acrónimo agencia financiadora AEI
Programa Programa Estatal para la Investigación y el Desarrollo Experimental
Subprograma Subprograma Estatal de Generación de Conocimiento Científico-Técnico y Desarrollo Experimental
Convocatoria Proyectos de I+D+i (Generación de Conocimiento y Retos Investigación)
Año convocatoria 2024
Plan Estatal 2024-2027 Plan Estatal de Investigación Científica y Técnica y de Innovación
Centro beneficiario UNIVERSIDAD POLITECNICA DE MADRID

Resultados relacionados

Resultados totales (Incluyendo duplicados): 1
Encontrada(s) 1 página(s)

DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Artículo científico (JournalArticle)

Towards safer chatbots: Automated policy compliance evaluation of custom GPTs

  • Rodriguez Torrado, David
  • Seymour, William
  • Álamo, Jose M. Del
  • Such, Jose
User-configured chatbots built on top of large language models are increasingly available through centralized marketplaces such as OpenAI’s GPT Store. While these platforms enforce usage policies intended to prevent harmful or inappropriate behavior, the scale and opacity of customized chatbots make systematic policy enforcement challenging. As a result, policy-violating chatbots continue to remain publicly accessible despite existing review processes. This paper presents a fully automated method for evaluating the compliance of Custom GPTs with its marketplace usage policy using black-box interaction. The method combines large-scale GPT discovery, policy-driven red-teaming prompts, and automated compliance assessment using an LLM-as-a-judge. We focus on three policy-relevant domains explicitly addressed in OpenAI’s usage policies: Romantic, Cybersecurity, and Academic GPTs. We validate our compliance assessment component against a human-annotated ground-truth dataset, achieving an F1 score of 0.975 for binary policy violation detection. We then apply the method in a large-scale empirical study of 782 Custom GPTs retrieved from the GPT Store. The results show that 58.7% of the evaluated GPTs exhibit at least one policy-violating response, with substantial variation across policy domains. A comparison with the base models (GPT-4 and GPT-4o) indicates that most violations originate from model-level behavior, with customization tending to amplify these tendencies rather than create new failure modes. Our findings reveal limitations in current review mechanisms for user-configured chatbots and demonstrate the feasibility of scalable, behavior-based policy compliance evaluation., The work of David Rodriguez was partially supported by the CEDAR project, funded by the Horizon Europe research program (2021–2027) under grant agreement no. 101135577. The work of Jose M. Del Alamo was partially supported by the Re-InitS project, funded by the Plan Estatal de Investigación Científica y Técnica y de Innovación 2024–2027 (Ministerio de Ciencia e Investigación (Spain) - MCIN/AEI/10.13039/501100011033) under grant agreement PID2024-155230OB-C43. The work of Jose Such was partially supported by Grant PID2023-151536OB-100 funded by MICIU/AEI/10.13039/501100011033 and by ERDF/EU, and by a PROMETEO 2024 (CIPROM/2023/23) grant of Conselleria d’Educació, Universitats i Ocupació (Generalitat Valenciana) ., Peer reviewed