LATIDIA · Ciberseguridad
Protección y auditoría de agentes LLM de largo plazo: auditoría y controles de transferencia de estado en el «harness» del agente
arXiv:2611.23346v1 Tipo de anuncio: nuevo Resumen: Los agentes LLM de largo alcance en bases de código, navegadores web e infraestructuras empresariales dependen de un «harness» de ejecución para trasladar el estado entre el contexto de trabajo, el historial de la sesión, etc.
WhatsApp ↗Telegram ↗
La noticia
arXiv:2611.23346v1 Announce Type: new Abstract: Long-horizon LLM agents across codebases, web browsers, and enterprise infrastructure rely on an execution harness to move state between the working context, session history, persistent memory, and external environment. While conventional execution traces log model prompts and tool calls, critical security and reliability failures occur during state transfers: a transfer can admit unchecked content (indirect prompt injections, unverified memory writes), rewrite state lossily (context compaction dropping user constraints, memory consolidation stripping source authority), or restore state incompletely (workspace rollbacks leaving aborted KV state active on the inference server). Because current traces omit what crosses these boundaries, failure attribution and runtime monitors fail on multi-step trajectories.