{"ok":true,"nota":{"id":"372c83abd639006d","slug":"un-informe-de-prueba-de-seguridad-detecto-que-openai-casi-no-vigilo-a--372c83","titulo":"Un informe de prueba de seguridad detectó que OpenAI casi no vigiló a sus agentes durante un ataque a Hugging ","resumen":"Un análisis de tipo red team concluyó que OpenAI tenía un monitoreo solo parcial sobre sus agentes de IA durante un ataque vinculado a Hugging Face. Además, ese control podía eludirse con facilidad.","fuente":"theregister","fuenteNombre":"The Register","color":"#ff0000","autor":"Thomas Claburn","url":"https://www.theregister.com/ai-and-ml/2026/10/23/red-team-report-found-openai-utterly-failed-to-monitor-its-agents-during-hugging-face-attack/5305200","imagen":"https://image.theregister.com/?imageId=5305204&width=800","tema":"tecnologia","subtemas":["inteligencia artificial","seguridad informatica","openai","agentes autonomos","hugging face"],"entidades":["OpenAI","Hugging Face","The Register"],"formato":"dato","entretenimiento":false,"pubTs":"2026-10-23T19:14:57.000Z","indexTs":"2026-10-23T22:31:31.923Z","hash":"38b5a6a03d3d08c019a893b4babbc3c13f8fee657cae273c2879f929d2c3ed51","tsa":{"time":"Oct 23 22:32:08 2026 GMT","provider":"freetsa.org","sobre":"sello"},"stats":{"views":0,"likes":0,"clicks":0,"dwell":0},"tipo":"motor","registro":"hallazgo","porQueImporta":"Las empresas de inteligencia artificial dicen que sus sistemas están bajo control, pero este informe muestra que la vigilancia sobre agentes autónomos puede ser bastante débil. Para cualquier persona o empresa de acá que empiece a usar estas herramientas para tareas reales, la pregunta es quién controla lo que hacen cuando nadie mira. El problema no es solo técnico: es de confianza en quien vende la tecnología.","quienFirma":"The Register (medio de tecnología), basado en un informe de red team","confianza":"divulgacion","consulta":{"pregunta":"¿Confiarías en un agente de IA que hace tareas sin supervisión?"}},"sello":{"hash":"38b5a6a03d3d08c019a893b4babbc3c13f8fee657cae273c2879f929d2c3ed51","sello":"a2174282153470bc5166eb00bfaba32728f6176a80118651ed03618267201913","beacon":{"drand":{"cadena":"8990e7a9aaed2ffed73dbd7092123d6f289930540d7651336225dc172e51b2ce","ronda":6578790,"aleatorio":"c4d4738914627cb599c4c0726095d0ca65d7e18b339aad7019f74f0cf22914d9","tiempo":"2026-10-23T22:32:00.000Z"},"nist":{"cadena":2,"pulso":1996588,"salida":"1AB33E4D4D8827C21F7F3557C5995F2A7803790AA889B11FBC3E8A1AAE593D51C4D6AF8D679624359D1D1868187DF78129DBF70BD659B8A94EB4847855F1EB6E","tiempo":"2026-10-23T22:31:00.000Z"},"noAntes":"2026-10-23T22:32:00.000Z"},"verificacion":{"drand":true,"nist":true},"tsa":{"time":"Oct 23 22:32:08 2026 GMT","provider":"freetsa.org","sobre":"sello"},"original":{"titulo":"Red-team report found OpenAI utterly failed to monitor its agents during Hugging Face attack","descripcion":"Monitoring was only partial and was easily bypassed","pubDate":"Fri, 23 Oct 2026 19:14:57 +0000","url":"https://www.theregister.com/ai-and-ml/2026/10/23/red-team-report-found-openai-utterly-failed-to-monitor-its-agents-during-hugging-face-attack/5305200"},"indexado":"2026-10-23T22:31:31.923Z"}}