LATIDIA · Ciberseguridad

Fallos correlacionados de los monitores de aprobación de IA

arXiv:2610.29178v1 Tipo de anuncio: nuevo Resumen: Antes de que un agente de IA tome una acción consecuente, como pagar, enviar mensajes o cambiar el código, varios monitores lo verifican. Sus veredictos se combinan mediante la suma de puntuaciones o

La noticia

arXiv:2610.29178v1 Announce Type: new Abstract: Before an AI agent takes a consequential action, such as paying, messaging or changing code, several monitors check it. Their verdicts are combined by aggregating scores or by counting approvals, whatever inputs the monitors share. Monitors that share a base model or a poisoned input, however, may fail together. We borrow the fail-prone systems of distributed computing: lists of monitor sets that may fail together. A rule is safe if it executes no bad action and live if it blocks no good one. We measured twelve small open-weights monitors on payment approval under two pre-registered protocols. Without pre-registration, we also measured three closed frontier models